OpenMed

--- widget: - text: "Administration of metformin reduced glucose levels significantly." - text: "The study evaluated the efficacy of cisplatin in cancer treatment." - text: "Patients received ibuprofen for inflammation management." - text: "The patient's medication was switched to tamoxifen to prevent breast cancer recurrence." - text: "Lithium carbonate is often prescribed for the management of bipolar disorder." tags: - token-classification - named-entity-recognition - biomedical-nlp - transfo

--- widget: - text: "The patient was administered acetylsalicylic acid for pain relief." - text: "Treatment with doxorubicin showed significant improvement in tumor regression." - text: "The compound benzylpenicillin demonstrated strong antimicrobial activity." - text: "Further studies are needed to understand the effects of methotrexate on rheumatoid arthritis." - text: "The synthesis of vancomycin remains a significant challenge in organic chemistry." tags: - token-classification - named-entit

--- widget: - text: "Administration of metformin reduced glucose levels significantly." - text: "The study evaluated the efficacy of cisplatin in cancer treatment." - text: "Patients received ibuprofen for inflammation management." - text: "The patient's medication was switched to tamoxifen to prevent breast cancer recurrence." - text: "Lithium carbonate is often prescribed for the management of bipolar disorder." tags: - token-classification - named-entity-recognition - biomedical-nlp - transfo

--- widget: - text: "The patient was administered acetylsalicylic acid for pain relief." - text: "Treatment with doxorubicin showed significant improvement in tumor regression." - text: "The compound benzylpenicillin demonstrated strong antimicrobial activity." - text: "Further studies are needed to understand the effects of methotrexate on rheumatoid arthritis." - text: "The synthesis of vancomycin remains a significant challenge in organic chemistry." tags: - token-classification - named-entit

--- widget: - text: "The patient was diagnosed with diabetes mellitus type 2." - text: "Symptoms of Alzheimer's disease became apparent over several months." - text: "Treatment for hypertension was initiated immediately." - text: "A possible link between Crohn's disease and gut microbiota is being investigated." - text: "The patient has a family history of cystic fibrosis." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - disease-entity-recognition - medi

--- widget: - text: "Mutations in KRAS gene drive oncogenic transformation." - text: "The tumor suppressor p53 pathway was disrupted." - text: "EGFR amplification promotes cancer cell proliferation." - text: "Loss of function of the PTEN gene is common in many cancers." - text: "The PI3K/AKT/mTOR pathway is a critical regulator of cell growth." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - cancer-genetics - oncology - gene-regulation - cancer-research

--- widget: - text: "Mutations in KRAS gene drive oncogenic transformation." - text: "The tumor suppressor p53 pathway was disrupted." - text: "EGFR amplification promotes cancer cell proliferation." - text: "Loss of function of the PTEN gene is common in many cancers." - text: "The PI3K/AKT/mTOR pathway is a critical regulator of cell growth." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - cancer-genetics - oncology - gene-regulation - cancer-research

--- widget: - text: "The BRCA2 gene is associated with hereditary breast cancer." - text: "Mutations in the CFTR gene cause cystic fibrosis." - text: "The APOE gene variant affects Alzheimer's disease risk." - text: "The HTT gene provides instructions for making a protein called huntingtin." - text: "Sickle cell disease is caused by a mutation in the HBB gene." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - gene-recognition - genetics - genomics - molec

--- widget: - text: "The patient complained of pain in the left ventricle region." - text: "Examination revealed inflammation of the hippocampus." - text: "The liver showed signs of fatty infiltration." - text: "An MRI of the cerebrum showed no signs of abnormalities." - text: "The procedure involved an incision near the femoral artery." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - anatomical-entity-recognition - medical-terminology - anatomy - health

--- widget: - text: "The patient was diagnosed with diabetes mellitus type 2." - text: "Symptoms of Alzheimer's disease became apparent over several months." - text: "Treatment for hypertension was initiated immediately." - text: "A possible link between Crohn's disease and gut microbiota is being investigated." - text: "The patient has a family history of cystic fibrosis." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - disease-entity-recognition - medi

--- widget: - text: "Escherichia coli bacteria were found in the water samples." - text: "The study included specimens from Homo sapiens and Mus musculus." - text: "Saccharomyces cerevisiae is commonly used in biotechnology applications." - text: "The venom of the black mamba, Dendroaspis polylepis, is highly neurotoxic." - text: "Canis lupus familiaris has been domesticated for thousands of years." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - species

--- widget: - text: "The EGFR gene mutation was identified in lung cancer patients." - text: "Overexpression of HER2 protein correlates with poor prognosis." - text: "The TP53 gene encodes a tumor suppressor protein." - text: "The BRAF V600E mutation is a common driver in melanoma." - text: "Insulin receptor signaling is essential for glucose homeostasis." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - gene-recognition - protein-recognition - genomics -

--- widget: - text: "The p53 protein plays a crucial role in tumor suppression." - text: "Expression of BRCA1 gene was significantly upregulated in breast tissue." - text: "The NF-kB pathway regulates inflammatory responses." - text: "Activation of the STAT3 signaling pathway is observed in many cancers." - text: "The experiment involved transfecting HeLa cells with a plasmid containing the target gene." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - pr

--- widget: - text: "Caenorhabditis elegans is a model organism for genetic studies." - text: "The research focused on Drosophila melanogaster development." - text: "Arabidopsis thaliana serves as a model for plant biology." - text: "The zebrafish, Danio rerio, is widely used for studying vertebrate development." - text: "Neurospora crassa is a type of red bread mold used in genetic research." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - species-recog

--- widget: - text: "The EGFR gene mutation was identified in lung cancer patients." - text: "Overexpression of HER2 protein correlates with poor prognosis." - text: "The TP53 gene encodes a tumor suppressor protein." - text: "The BRAF V600E mutation is a common driver in melanoma." - text: "Insulin receptor signaling is essential for glucose homeostasis." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - gene-recognition - protein-recognition - genomics -

--- widget: - text: "Early detection of breast cancer improves survival rates." - text: "The patient exhibited symptoms consistent with Parkinson's disease." - text: "Genetic testing revealed predisposition to Huntington's disease." - text: "Malaria is a life-threatening disease caused by parasites transmitted through mosquito bites." - text: "Multiple sclerosis affects the central nervous system, leading to a range of symptoms." tags: - token-classification - named-entity-recognition - biomedic

--- widget: - text: "The patient was administered acetylsalicylic acid for pain relief." - text: "Treatment with doxorubicin showed significant improvement in tumor regression." - text: "The compound benzylpenicillin demonstrated strong antimicrobial activity." - text: "Further studies are needed to understand the effects of methotrexate on rheumatoid arthritis." - text: "The synthesis of vancomycin remains a significant challenge in organic chemistry." tags: - token-classification - named-entit

--- widget: - text: "The p53 protein plays a crucial role in tumor suppression." - text: "Expression of BRCA1 gene was significantly upregulated in breast tissue." - text: "The NF-kB pathway regulates inflammatory responses." - text: "Activation of the STAT3 signaling pathway is observed in many cancers." - text: "The experiment involved transfecting HeLa cells with a plasmid containing the target gene." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - pr

--- widget: - text: "Escherichia coli bacteria were found in the water samples." - text: "The study included specimens from Homo sapiens and Mus musculus." - text: "Saccharomyces cerevisiae is commonly used in biotechnology applications." - text: "The venom of the black mamba, Dendroaspis polylepis, is highly neurotoxic." - text: "Canis lupus familiaris has been domesticated for thousands of years." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - species

--- widget: - text: "Administration of metformin reduced glucose levels significantly." - text: "The study evaluated the efficacy of cisplatin in cancer treatment." - text: "Patients received ibuprofen for inflammation management." - text: "The patient's medication was switched to tamoxifen to prevent breast cancer recurrence." - text: "Lithium carbonate is often prescribed for the management of bipolar disorder." tags: - token-classification - named-entity-recognition - biomedical-nlp - transfo

--- widget: - text: "Caenorhabditis elegans is a model organism for genetic studies." - text: "The research focused on Drosophila melanogaster development." - text: "Arabidopsis thaliana serves as a model for plant biology." - text: "The zebrafish, Danio rerio, is widely used for studying vertebrate development." - text: "Neurospora crassa is a type of red bread mold used in genetic research." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - species-recog

--- widget: - text: "The patient presented with chronic lymphocytic leukemia symptoms." - text: "B-cell proliferation was observed in bone marrow samples." - text: "Treatment with ibrutinib showed promising results." - text: "Flow cytometry confirmed the diagnosis of chronic lymphocytic leukemia." - text: "The patient had del(17p), a high-risk feature in CLL." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - leukemia - hematology - cancer - clinical-medic

--- widget: - text: "The BRCA2 gene is associated with hereditary breast cancer." - text: "Mutations in the CFTR gene cause cystic fibrosis." - text: "The APOE gene variant affects Alzheimer's disease risk." - text: "The HTT gene provides instructions for making a protein called huntingtin." - text: "Sickle cell disease is caused by a mutation in the HBB gene." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - gene-recognition - genetics - genomics - molec

--- widget: - text: "The patient was administered acetylsalicylic acid for pain relief." - text: "Treatment with doxorubicin showed significant improvement in tumor regression." - text: "The compound benzylpenicillin demonstrated strong antimicrobial activity." - text: "Further studies are needed to understand the effects of methotrexate on rheumatoid arthritis." - text: "The synthesis of vancomycin remains a significant challenge in organic chemistry." tags: - token-classification - named-entit

--- widget: - text: "The Maillard reaction is responsible for the browning of many foods." - text: "Casein micelles are the primary protein component of milk." - text: "Starch gelatinization is a key process in cooking pasta and rice." - text: "Polyphenols in green tea have antioxidant properties." - text: "Omega-3 fatty acids are essential fats found in fish oil." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - protein-interactions - molecular-biology -

--- widget: - text: "The patient was administered acetylsalicylic acid for pain relief." - text: "Treatment with doxorubicin showed significant improvement in tumor regression." - text: "The compound benzylpenicillin demonstrated strong antimicrobial activity." - text: "Further studies are needed to understand the effects of methotrexate on rheumatoid arthritis." - text: "The synthesis of vancomycin remains a significant challenge in organic chemistry." tags: - token-classification - named-entit

--- widget: - text: "Escherichia coli bacteria were found in the water samples." - text: "The study included specimens from Homo sapiens and Mus musculus." - text: "Saccharomyces cerevisiae is commonly used in biotechnology applications." - text: "The venom of the black mamba, Dendroaspis polylepis, is highly neurotoxic." - text: "Canis lupus familiaris has been domesticated for thousands of years." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - species

--- widget: - text: "The patient presented with chronic lymphocytic leukemia symptoms." - text: "B-cell proliferation was observed in bone marrow samples." - text: "Treatment with ibrutinib showed promising results." - text: "Flow cytometry confirmed the diagnosis of chronic lymphocytic leukemia." - text: "The patient had del(17p), a high-risk feature in CLL." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - leukemia - hematology - cancer - clinical-medic

--- widget: - text: "Administration of metformin reduced glucose levels significantly." - text: "The study evaluated the efficacy of cisplatin in cancer treatment." - text: "Patients received ibuprofen for inflammation management." - text: "The patient's medication was switched to tamoxifen to prevent breast cancer recurrence." - text: "Lithium carbonate is often prescribed for the management of bipolar disorder." tags: - token-classification - named-entity-recognition - biomedical-nlp - transfo

--- widget: - text: "The patient was administered acetylsalicylic acid for pain relief." - text: "Treatment with doxorubicin showed significant improvement in tumor regression." - text: "The compound benzylpenicillin demonstrated strong antimicrobial activity." - text: "Further studies are needed to understand the effects of methotrexate on rheumatoid arthritis." - text: "The synthesis of vancomycin remains a significant challenge in organic chemistry." tags: - token-classification - named-entit

--- widget: - text: "Caenorhabditis elegans is a model organism for genetic studies." - text: "The research focused on Drosophila melanogaster development." - text: "Arabidopsis thaliana serves as a model for plant biology." - text: "The zebrafish, Danio rerio, is widely used for studying vertebrate development." - text: "Neurospora crassa is a type of red bread mold used in genetic research." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - species-recog

--- widget: - text: "The BRCA2 gene is associated with hereditary breast cancer." - text: "Mutations in the CFTR gene cause cystic fibrosis." - text: "The APOE gene variant affects Alzheimer's disease risk." - text: "The HTT gene provides instructions for making a protein called huntingtin." - text: "Sickle cell disease is caused by a mutation in the HBB gene." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - gene-recognition - genetics - genomics - molec

--- widget: - text: "The patient was diagnosed with diabetes mellitus type 2." - text: "Symptoms of Alzheimer's disease became apparent over several months." - text: "Treatment for hypertension was initiated immediately." - text: "A possible link between Crohn's disease and gut microbiota is being investigated." - text: "The patient has a family history of cystic fibrosis." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - disease-entity-recognition - medi

--- widget: - text: "The BRCA2 gene is associated with hereditary breast cancer." - text: "Mutations in the CFTR gene cause cystic fibrosis." - text: "The APOE gene variant affects Alzheimer's disease risk." - text: "The HTT gene provides instructions for making a protein called huntingtin." - text: "Sickle cell disease is caused by a mutation in the HBB gene." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - gene-recognition - genetics - genomics - molec

--- widget: - text: "Mutations in KRAS gene drive oncogenic transformation." - text: "The tumor suppressor p53 pathway was disrupted." - text: "EGFR amplification promotes cancer cell proliferation." - text: "Loss of function of the PTEN gene is common in many cancers." - text: "The PI3K/AKT/mTOR pathway is a critical regulator of cell growth." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - cancer-genetics - oncology - gene-regulation - cancer-research

--- widget: - text: "Early detection of breast cancer improves survival rates." - text: "The patient exhibited symptoms consistent with Parkinson's disease." - text: "Genetic testing revealed predisposition to Huntington's disease." - text: "Malaria is a life-threatening disease caused by parasites transmitted through mosquito bites." - text: "Multiple sclerosis affects the central nervous system, leading to a range of symptoms." tags: - token-classification - named-entity-recognition - biomedic

--- widget: - text: "Escherichia coli bacteria were found in the water samples." - text: "The study included specimens from Homo sapiens and Mus musculus." - text: "Saccharomyces cerevisiae is commonly used in biotechnology applications." - text: "The venom of the black mamba, Dendroaspis polylepis, is highly neurotoxic." - text: "Canis lupus familiaris has been domesticated for thousands of years." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - species

--- widget: - text: "The patient was diagnosed with diabetes mellitus type 2." - text: "Symptoms of Alzheimer's disease became apparent over several months." - text: "Treatment for hypertension was initiated immediately." - text: "A possible link between Crohn's disease and gut microbiota is being investigated." - text: "The patient has a family history of cystic fibrosis." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - disease-entity-recognition - medi

--- widget: - text: "The EGFR gene mutation was identified in lung cancer patients." - text: "Overexpression of HER2 protein correlates with poor prognosis." - text: "The TP53 gene encodes a tumor suppressor protein." - text: "The BRAF V600E mutation is a common driver in melanoma." - text: "Insulin receptor signaling is essential for glucose homeostasis." tags: - token-classification - named-entity-recognition - biomedical-nlp - transformers - gene-recognition - protein-recognition - genomics -

--- widget: - text: "Administration of metformin reduced glucose levels significantly." - text: "The study evaluated the efficacy of cisplatin in cancer treatment." - text: "Patients received ibuprofen for inflammation management." - text: "The patient's medication was switched to tamoxifen to prevent breast cancer recurrence." - text: "Lithium carbonate is often prescribed for the management of bipolar disorder." tags: - token-classification - named-entity-recognition - biomedical-nlp - transfo

--- widget: - text: "Early detection of breast cancer improves survival rates." - text: "The patient exhibited symptoms consistent with Parkinson's disease." - text: "Genetic testing revealed predisposition to Huntington's disease." - text: "Malaria is a life-threatening disease caused by parasites transmitted through mosquito bites." - text: "Multiple sclerosis affects the central nervous system, leading to a range of symptoms." tags: - token-classification - named-entity-recognition - biomedic

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.90` - Precision: `0.89` - Recall: `0.91` - Accuracy: `0.94` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.90` - Precision: `0.88` - Recall: `0.92` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.88` - Precision: `0.88` - Recall: `0.87` - Accuracy: `0.95` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.90` - Precision: `0.90` - Recall: `0.91` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: deberta-v3-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.82` - Precision: `0.78` - Recall: `0.86` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: deberta-v3-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.91` - Precision: `0.92` - Recall: `0.91` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: EuroBERT-210m - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: EuroBERT-210m - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.81` - Precision: `0.80` - Recall: `0.81` - Accuracy: `0.94` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species names from the Species-800 dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species names from the species-800 dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated SPECIES800 dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Species800 is a corpus for species recognition and taxonomy classification in biomedical texts. The Species800 corpus is a manually annotated dataset designed for species recognition and taxonomic classification in biomedical literature. This corpus contains 800 abstracts with comprehensive annotations for organism mentions, supporting biodiversity informatics and biological taxonomy research. The dataset includes both scientific names and common names of species, making it valuable for developing NER systems that can handle the complexity of biological nomenclature. It serves as a benchmark for evaluating species identification models used in ecological studies, conservation biology, and systematic biology research. The corpus is particularly useful for text mining applications in biodiversity databases and biological literature analysis. Current Model Performance - F1 Score: `0.86` - Precision: `0.86` - Recall: `0.87` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OrganismDetect-BioMed-335M | 0.8639 | 0.8557 | 0.8722 | 0.9715 | | 🥈 2 | OpenMed-NER-OrganismDetect-PubMed-335M | 0.8550 | 0.8370 | 0.8737 | 0.9698 | | 🥉 3 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8458 | 0.8287 | 0.8637 | 0.9690 | | 4 | OpenMed-NER-OrganismDetect-MultiMed-335M | 0.8441 | 0.8352 | 0.8532 | 0.9670 | | 5 | OpenMed-NER-OrganismDetect-SuperClinical-434M | 0.8435 | 0.8291 | 0.8585 | 0.9670 | | 6 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8349 | 0.8082 | 0.8634 | 0.9685 | | 7 | OpenMed-NER-OrganismDetect-MultiMed-568M | 0.8313 | 0.8053 | 0.8592 | 0.9703 | | 8 | OpenMed-NER-OrganismDetect-ElectraMed-335M | 0.8288 | 0.8176 | 0.8404 | 0.9631 | | 9 | OpenMed-NER-OrganismDetect-BioPatient-108M | 0.8154 | 0.8140 | 0.8169 | 0.9591 | | 10 | OpenMed-NER-OrganismDetect-ElectraMed-33M | 0.8121 | 0.7772 | 0.8503 | 0.9600 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: SPECIES800 - Description: Species Entity Recognition - Species names from the Species-800 dataset Training Details - Base Model: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Anatomical Entity Recognition - Anatomical structures and body parts [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for anatomical entity recognition - anatomical structures and body parts. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated ANATOMY dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Anatomy corpus focuses on anatomical entity recognition for medical terminology and healthcare applications. The Anatomy corpus is a specialized biomedical NER dataset designed for recognizing anatomical entities and medical terminology in clinical and biomedical texts. This corpus contains annotations for anatomical structures, body parts, organs, and physiological systems mentioned in medical literature. It is essential for developing clinical NLP systems, medical education tools, and healthcare informatics applications where accurate anatomical entity identification is crucial. The dataset supports the development of automated systems for medical coding, clinical decision support, and anatomical knowledge extraction from medical records and literature. It serves as a valuable resource for training NER models used in medical imaging, surgical planning, and clinical documentation. Current Model Performance - F1 Score: `0.91` - Precision: `0.90` - Recall: `0.91` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-AnatomyDetect-ElectraMed-560M | 0.9063 | 0.9083 | 0.9044 | 0.9825 | | 🥈 2 | OpenMed-NER-AnatomyDetect-PubMed-335M | 0.9063 | 0.8995 | 0.9131 | 0.9851 | | 🥉 3 | OpenMed-NER-AnatomyDetect-SuperClinical-434M | 0.9024 | 0.9040 | 0.9008 | 0.9836 | | 4 | OpenMed-NER-AnatomyDetect-ElectraMed-335M | 0.9020 | 0.9024 | 0.9016 | 0.9787 | | 5 | OpenMed-NER-AnatomyDetect-MultiMed-568M | 0.9012 | 0.8977 | 0.9048 | 0.9812 | | 6 | OpenMed-NER-AnatomyDetect-PubMed-109M | 0.9004 | 0.8941 | 0.9067 | 0.9844 | | 7 | OpenMed-NER-AnatomyDetect-SuperMedical-355M | 0.9002 | 0.8974 | 0.9029 | 0.9815 | | 8 | OpenMed-NER-AnatomyDetect-BigMed-560M | 0.8980 | 0.9007 | 0.8954 | 0.9814 | | 9 | OpenMed-NER-AnatomyDetect-BioMed-335M | 0.8961 | 0.8941 | 0.8982 | 0.9830 | | 10 | OpenMed-NER-AnatomyDetect-BioClinical-108M | 0.8961 | 0.8960 | 0.8962 | 0.9768 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: ANATOMY - Description: Anatomical Entity Recognition - Anatomical structures and body parts Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

🧬 OpenMed-NER-BloodCancerDetect-SuperClinical-434M Specialized model for Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for clinical entity recognition - clinical entities related to chronic lymphocytic leukemia. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated CLL dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: CLL corpus is specialized for chronic lymphocytic leukemia entity recognition in hematology and cancer research. The CLL (Chronic Lymphocytic Leukemia) corpus is a domain-specific biomedical NER dataset focused on entities related to chronic lymphocytic leukemia, a type of blood cancer. This specialized corpus contains annotations for CLL-specific terminology, biomarkers, treatment entities, and clinical concepts relevant to hematology and oncology research. The dataset is designed to support the development of clinical NLP systems for leukemia research, hematological disorder analysis, and cancer informatics applications. It is particularly valuable for identifying disease-specific entities, therapeutic interventions, and prognostic factors mentioned in CLL research literature. The corpus serves as a benchmark for evaluating NER models in specialized medical domains and clinical research. Current Model Performance - F1 Score: `0.89` - Precision: `0.87` - Recall: `0.92` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-BloodCancerDetect-ElectraMed-560M | 0.9575 | 0.9264 | 0.9907 | 0.9843 | | 🥈 2 | OpenMed-NER-BloodCancerDetect-SuperClinical-434M | 0.8902 | 0.8652 | 0.9167 | 0.9701 | | 🥉 3 | OpenMed-NER-BloodCancerDetect-TinyMed-82M | 0.8793 | 0.7904 | 0.9908 | 0.9449 | | 4 | OpenMed-NER-BloodCancerDetect-TinyMed-135M | 0.8792 | 0.8750 | 0.8835 | 0.9668 | | 5 | OpenMed-NER-BloodCancerDetect-TinyMed-65M | 0.8547 | 0.7812 | 0.9434 | 0.9686 | | 6 | OpenMed-NER-BloodCancerDetect-SuperMedical-125M | 0.8488 | 1.0000 | 0.7373 | 0.9274 | | 7 | OpenMed-NER-BloodCancerDetect-SnowMed-568M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 8 | OpenMed-NER-BloodCancerDetect-BigMed-278M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 9 | OpenMed-NER-BloodCancerDetect-SuperMedical-355M | 0.8421 | 0.9816 | 0.7373 | 0.9248 | | 10 | OpenMed-NER-BloodCancerDetect-ElectraMed-335M | 0.8364 | 0.7302 | 0.9787 | 0.9581 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: CLL - Description: Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia Training Details - Base Model: deberta-v3-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.96` - Precision: `0.95` - Recall: `0.97` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.91` - Precision: `0.89` - Recall: `0.93` - Accuracy: `0.98` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.96` - Precision: `0.95` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.95` - Precision: `0.94` - Recall: `0.95` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.94` - Precision: `0.93` - Recall: `0.95` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.68` - Precision: `0.66` - Recall: `0.69` - Accuracy: `0.92` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: EuroBERT-210m - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: EuroBERT-210m - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.95` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: EuroBERT-210m - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: EuroBERT-210m - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.86` - Precision: `0.86` - Recall: `0.85` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: xlm-roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.96` - Precision: `0.96` - Recall: `0.96` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.87` - Precision: `0.86` - Recall: `0.89` - Accuracy: `0.98` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.95` - Precision: `0.94` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: ModernBERT-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: ModernBERT-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Anatomical Entity Recognition - Anatomical structures and body parts [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for anatomical entity recognition - anatomical structures and body parts. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated ANATOMY dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Anatomy corpus focuses on anatomical entity recognition for medical terminology and healthcare applications. The Anatomy corpus is a specialized biomedical NER dataset designed for recognizing anatomical entities and medical terminology in clinical and biomedical texts. This corpus contains annotations for anatomical structures, body parts, organs, and physiological systems mentioned in medical literature. It is essential for developing clinical NLP systems, medical education tools, and healthcare informatics applications where accurate anatomical entity identification is crucial. The dataset supports the development of automated systems for medical coding, clinical decision support, and anatomical knowledge extraction from medical records and literature. It serves as a valuable resource for training NER models used in medical imaging, surgical planning, and clinical documentation. Current Model Performance - F1 Score: `0.87` - Precision: `0.87` - Recall: `0.87` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-AnatomyDetect-ElectraMed-560M | 0.9063 | 0.9083 | 0.9044 | 0.9825 | | 🥈 2 | OpenMed-NER-AnatomyDetect-PubMed-335M | 0.9063 | 0.8995 | 0.9131 | 0.9851 | | 🥉 3 | OpenMed-NER-AnatomyDetect-SuperClinical-434M | 0.9024 | 0.9040 | 0.9008 | 0.9836 | | 4 | OpenMed-NER-AnatomyDetect-ElectraMed-335M | 0.9020 | 0.9024 | 0.9016 | 0.9787 | | 5 | OpenMed-NER-AnatomyDetect-MultiMed-568M | 0.9012 | 0.8977 | 0.9048 | 0.9812 | | 6 | OpenMed-NER-AnatomyDetect-PubMed-109M | 0.9004 | 0.8941 | 0.9067 | 0.9844 | | 7 | OpenMed-NER-AnatomyDetect-SuperMedical-355M | 0.9002 | 0.8974 | 0.9029 | 0.9815 | | 8 | OpenMed-NER-AnatomyDetect-BigMed-560M | 0.8980 | 0.9007 | 0.8954 | 0.9814 | | 9 | OpenMed-NER-AnatomyDetect-BioMed-335M | 0.8961 | 0.8941 | 0.8982 | 0.9830 | | 10 | OpenMed-NER-AnatomyDetect-BioClinical-108M | 0.8961 | 0.8960 | 0.8962 | 0.9768 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: ANATOMY - Description: Anatomical Entity Recognition - Anatomical structures and body parts Training Details - Base Model: distilbert-base-multilingual-cased - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: distilbert-base-multilingual-cased - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.88` - Precision: `0.87` - Recall: `0.89` - Accuracy: `0.97` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: BioClinical-ModernBERT-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Anatomical Entity Recognition - Anatomical structures and body parts [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for anatomical entity recognition - anatomical structures and body parts. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated ANATOMY dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Anatomy corpus focuses on anatomical entity recognition for medical terminology and healthcare applications. The Anatomy corpus is a specialized biomedical NER dataset designed for recognizing anatomical entities and medical terminology in clinical and biomedical texts. This corpus contains annotations for anatomical structures, body parts, organs, and physiological systems mentioned in medical literature. It is essential for developing clinical NLP systems, medical education tools, and healthcare informatics applications where accurate anatomical entity identification is crucial. The dataset supports the development of automated systems for medical coding, clinical decision support, and anatomical knowledge extraction from medical records and literature. It serves as a valuable resource for training NER models used in medical imaging, surgical planning, and clinical documentation. Current Model Performance - F1 Score: `0.90` - Precision: `0.90` - Recall: `0.90` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-AnatomyDetect-ElectraMed-560M | 0.9063 | 0.9083 | 0.9044 | 0.9825 | | 🥈 2 | OpenMed-NER-AnatomyDetect-PubMed-335M | 0.9063 | 0.8995 | 0.9131 | 0.9851 | | 🥉 3 | OpenMed-NER-AnatomyDetect-SuperClinical-434M | 0.9024 | 0.9040 | 0.9008 | 0.9836 | | 4 | OpenMed-NER-AnatomyDetect-ElectraMed-335M | 0.9020 | 0.9024 | 0.9016 | 0.9787 | | 5 | OpenMed-NER-AnatomyDetect-MultiMed-568M | 0.9012 | 0.8977 | 0.9048 | 0.9812 | | 6 | OpenMed-NER-AnatomyDetect-PubMed-109M | 0.9004 | 0.8941 | 0.9067 | 0.9844 | | 7 | OpenMed-NER-AnatomyDetect-SuperMedical-355M | 0.9002 | 0.8974 | 0.9029 | 0.9815 | | 8 | OpenMed-NER-AnatomyDetect-BigMed-560M | 0.8980 | 0.9007 | 0.8954 | 0.9814 | | 9 | OpenMed-NER-AnatomyDetect-BioMed-335M | 0.8961 | 0.8941 | 0.8982 | 0.9830 | | 10 | OpenMed-NER-AnatomyDetect-BioClinical-108M | 0.8961 | 0.8960 | 0.8962 | 0.9768 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: ANATOMY - Description: Anatomical Entity Recognition - Anatomical structures and body parts Training Details - Base Model: deberta-v3-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.89` - Precision: `0.87` - Recall: `0.92` - Accuracy: `0.98` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: deberta-v3-small - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-small - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.89` - Precision: `0.86` - Recall: `0.92` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: deberta-v3-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.97` - Precision: `1.00` - Recall: `0.95` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: BioClinical-ModernBERT-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.96` - Precision: `0.94` - Recall: `0.97` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.90` - Precision: `0.89` - Recall: `0.91` - Accuracy: `0.97` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: BioDischargeSummaryBERT - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioDischargeSummaryBERT - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.95` - Precision: `0.94` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: bge-m3 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-m3 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.96` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.90` - Precision: `0.89` - Recall: `0.92` - Accuracy: `0.98` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: bge-large-en-v1.5 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-large-en-v1.5 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.90` - Precision: `0.89` - Recall: `0.91` - Accuracy: `0.98` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.95` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.91` - Precision: `0.89` - Recall: `0.93` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.95` - Precision: `0.94` - Recall: `0.95` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.90` - Precision: `0.87` - Recall: `0.92` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.99` - Precision: `0.99` - Recall: `0.99` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.88` - Precision: `0.87` - Recall: `0.90` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: BioClinical-ModernBERT-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.73` - Precision: `0.72` - Recall: `0.74` - Accuracy: `0.90` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: e5-base-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-base-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.89` - Precision: `0.87` - Recall: `0.91` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species and organism names [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species and organism names. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated LINNAEUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Linnaeus corpus is designed for species name identification and taxonomic entity recognition in biomedical literature. The Linnaeus corpus is a specialized biomedical NER dataset focused on species name identification and organism recognition in scientific literature. Named after Carl Linnaeus who established modern taxonomic nomenclature, this corpus contains annotations for species mentions that are normalized to NCBI Taxonomy identifiers. The dataset is crucial for biodiversity informatics, ecological research, and biological literature mining where accurate organism identification is essential. It supports the development of text mining systems for taxonomic studies, species distribution research, and comparative genomics applications. The corpus addresses the challenge of recognizing both scientific names and common names of organisms across diverse biological texts. Current Model Performance - F1 Score: `0.76` - Precision: `0.72` - Recall: `0.82` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-SpeciesDetect-PubMed-335M | 0.9649 | 0.9582 | 0.9718 | 0.9967 | | 🥈 2 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9543 | 0.9422 | 0.9667 | 0.9956 | | 🥉 3 | OpenMed-NER-SpeciesDetect-BioMed-335M | 0.9539 | 0.9441 | 0.9638 | 0.9957 | | 4 | OpenMed-NER-SpeciesDetect-SuperClinical-434M | 0.9534 | 0.9369 | 0.9704 | 0.9959 | | 5 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9502 | 0.9317 | 0.9695 | 0.9951 | | 6 | OpenMed-NER-SpeciesDetect-MultiMed-335M | 0.9479 | 0.9286 | 0.9680 | 0.9955 | | 7 | OpenMed-NER-SpeciesDetect-MultiMed-568M | 0.9460 | 0.9312 | 0.9613 | 0.9957 | | 8 | OpenMed-NER-SpeciesDetect-SuperMedical-355M | 0.9433 | 0.9221 | 0.9655 | 0.9953 | | 9 | OpenMed-NER-SpeciesDetect-SuperClinical-141M | 0.9406 | 0.9290 | 0.9525 | 0.9950 | | 10 | OpenMed-NER-SpeciesDetect-ModernClinical-395M | 0.9385 | 0.9379 | 0.9392 | 0.9940 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: LINNAEUS - Description: Species Entity Recognition - Species and organism names Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.82` - Precision: `0.79` - Recall: `0.85` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: e5-small-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-small-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species names from the Species-800 dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species names from the species-800 dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated SPECIES800 dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Species800 is a corpus for species recognition and taxonomy classification in biomedical texts. The Species800 corpus is a manually annotated dataset designed for species recognition and taxonomic classification in biomedical literature. This corpus contains 800 abstracts with comprehensive annotations for organism mentions, supporting biodiversity informatics and biological taxonomy research. The dataset includes both scientific names and common names of species, making it valuable for developing NER systems that can handle the complexity of biological nomenclature. It serves as a benchmark for evaluating species identification models used in ecological studies, conservation biology, and systematic biology research. The corpus is particularly useful for text mining applications in biodiversity databases and biological literature analysis. Current Model Performance - F1 Score: `0.85` - Precision: `0.84` - Recall: `0.87` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OrganismDetect-BioMed-335M | 0.8639 | 0.8557 | 0.8722 | 0.9715 | | 🥈 2 | OpenMed-NER-OrganismDetect-PubMed-335M | 0.8550 | 0.8370 | 0.8737 | 0.9698 | | 🥉 3 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8458 | 0.8287 | 0.8637 | 0.9690 | | 4 | OpenMed-NER-OrganismDetect-MultiMed-335M | 0.8441 | 0.8352 | 0.8532 | 0.9670 | | 5 | OpenMed-NER-OrganismDetect-SuperClinical-434M | 0.8435 | 0.8291 | 0.8585 | 0.9670 | | 6 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8349 | 0.8082 | 0.8634 | 0.9685 | | 7 | OpenMed-NER-OrganismDetect-MultiMed-568M | 0.8313 | 0.8053 | 0.8592 | 0.9703 | | 8 | OpenMed-NER-OrganismDetect-ElectraMed-335M | 0.8288 | 0.8176 | 0.8404 | 0.9631 | | 9 | OpenMed-NER-OrganismDetect-BioPatient-108M | 0.8154 | 0.8140 | 0.8169 | 0.9591 | | 10 | OpenMed-NER-OrganismDetect-ElectraMed-33M | 0.8121 | 0.7772 | 0.8503 | 0.9600 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: SPECIES800 - Description: Species Entity Recognition - Species names from the Species-800 dataset Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.95` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.93` - Precision: `0.92` - Recall: `0.94` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: BioClinical-ModernBERT-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.85` - Precision: `0.84` - Recall: `0.86` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: BioClinical-ModernBERT-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.94` - Precision: `0.94` - Recall: `0.95` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.89` - Precision: `0.86` - Recall: `0.92` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: deberta-v3-small - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-small - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.99` - Precision: `1.00` - Recall: `0.99` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.87` - Precision: `0.86` - Recall: `0.88` - Accuracy: `0.97` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `1.00` - Precision: `1.00` - Recall: `1.00` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.88` - Precision: `0.86` - Recall: `0.89` - Accuracy: `0.98` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: roberta-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: roberta-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.87` - Precision: `0.86` - Recall: `0.88` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.99` - Precision: `0.99` - Recall: `0.99` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: BioDischargeSummaryBERT - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioDischargeSummaryBERT - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.91` - Precision: `0.90` - Recall: `0.92` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: distilbert-base-multilingual-cased - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: distilbert-base-multilingual-cased - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.77` - Precision: `0.73` - Recall: `0.83` - Accuracy: `0.90` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: e5-base-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-base-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Anatomical Entity Recognition - Anatomical structures and body parts [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for anatomical entity recognition - anatomical structures and body parts. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated ANATOMY dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Anatomy corpus focuses on anatomical entity recognition for medical terminology and healthcare applications. The Anatomy corpus is a specialized biomedical NER dataset designed for recognizing anatomical entities and medical terminology in clinical and biomedical texts. This corpus contains annotations for anatomical structures, body parts, organs, and physiological systems mentioned in medical literature. It is essential for developing clinical NLP systems, medical education tools, and healthcare informatics applications where accurate anatomical entity identification is crucial. The dataset supports the development of automated systems for medical coding, clinical decision support, and anatomical knowledge extraction from medical records and literature. It serves as a valuable resource for training NER models used in medical imaging, surgical planning, and clinical documentation. Current Model Performance - F1 Score: `0.86` - Precision: `0.86` - Recall: `0.85` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-AnatomyDetect-ElectraMed-560M | 0.9063 | 0.9083 | 0.9044 | 0.9825 | | 🥈 2 | OpenMed-NER-AnatomyDetect-PubMed-335M | 0.9063 | 0.8995 | 0.9131 | 0.9851 | | 🥉 3 | OpenMed-NER-AnatomyDetect-SuperClinical-434M | 0.9024 | 0.9040 | 0.9008 | 0.9836 | | 4 | OpenMed-NER-AnatomyDetect-ElectraMed-335M | 0.9020 | 0.9024 | 0.9016 | 0.9787 | | 5 | OpenMed-NER-AnatomyDetect-MultiMed-568M | 0.9012 | 0.8977 | 0.9048 | 0.9812 | | 6 | OpenMed-NER-AnatomyDetect-PubMed-109M | 0.9004 | 0.8941 | 0.9067 | 0.9844 | | 7 | OpenMed-NER-AnatomyDetect-SuperMedical-355M | 0.9002 | 0.8974 | 0.9029 | 0.9815 | | 8 | OpenMed-NER-AnatomyDetect-BigMed-560M | 0.8980 | 0.9007 | 0.8954 | 0.9814 | | 9 | OpenMed-NER-AnatomyDetect-BioMed-335M | 0.8961 | 0.8941 | 0.8982 | 0.9830 | | 10 | OpenMed-NER-AnatomyDetect-BioClinical-108M | 0.8961 | 0.8960 | 0.8962 | 0.9768 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: ANATOMY - Description: Anatomical Entity Recognition - Anatomical structures and body parts Training Details - Base Model: BioClinical-ModernBERT-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for clinical entity recognition - clinical entities related to chronic lymphocytic leukemia. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated CLL dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: CLL corpus is specialized for chronic lymphocytic leukemia entity recognition in hematology and cancer research. The CLL (Chronic Lymphocytic Leukemia) corpus is a domain-specific biomedical NER dataset focused on entities related to chronic lymphocytic leukemia, a type of blood cancer. This specialized corpus contains annotations for CLL-specific terminology, biomarkers, treatment entities, and clinical concepts relevant to hematology and oncology research. The dataset is designed to support the development of clinical NLP systems for leukemia research, hematological disorder analysis, and cancer informatics applications. It is particularly valuable for identifying disease-specific entities, therapeutic interventions, and prognostic factors mentioned in CLL research literature. The corpus serves as a benchmark for evaluating NER models in specialized medical domains and clinical research. Current Model Performance - F1 Score: `0.53` - Precision: `0.48` - Recall: `0.60` - Accuracy: `0.91` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-BloodCancerDetect-ElectraMed-560M | 0.9575 | 0.9264 | 0.9907 | 0.9843 | | 🥈 2 | OpenMed-NER-BloodCancerDetect-SuperClinical-434M | 0.8902 | 0.8652 | 0.9167 | 0.9701 | | 🥉 3 | OpenMed-NER-BloodCancerDetect-TinyMed-82M | 0.8793 | 0.7904 | 0.9908 | 0.9449 | | 4 | OpenMed-NER-BloodCancerDetect-TinyMed-135M | 0.8792 | 0.8750 | 0.8835 | 0.9668 | | 5 | OpenMed-NER-BloodCancerDetect-TinyMed-65M | 0.8547 | 0.7812 | 0.9434 | 0.9686 | | 6 | OpenMed-NER-BloodCancerDetect-SuperMedical-125M | 0.8488 | 1.0000 | 0.7373 | 0.9274 | | 7 | OpenMed-NER-BloodCancerDetect-SnowMed-568M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 8 | OpenMed-NER-BloodCancerDetect-BigMed-278M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 9 | OpenMed-NER-BloodCancerDetect-SuperMedical-355M | 0.8421 | 0.9816 | 0.7373 | 0.9248 | | 10 | OpenMed-NER-BloodCancerDetect-ElectraMed-335M | 0.8364 | 0.7302 | 0.9787 | 0.9581 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: CLL - Description: Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.79` - Precision: `0.74` - Recall: `0.85` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.96` - Precision: `0.96` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: bge-large-en-v1.5 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-large-en-v1.5 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.79` - Precision: `0.75` - Recall: `0.84` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.94` - Precision: `0.93` - Recall: `0.95` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: e5-small-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-small-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species and organism names [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species and organism names. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated LINNAEUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Linnaeus corpus is designed for species name identification and taxonomic entity recognition in biomedical literature. The Linnaeus corpus is a specialized biomedical NER dataset focused on species name identification and organism recognition in scientific literature. Named after Carl Linnaeus who established modern taxonomic nomenclature, this corpus contains annotations for species mentions that are normalized to NCBI Taxonomy identifiers. The dataset is crucial for biodiversity informatics, ecological research, and biological literature mining where accurate organism identification is essential. It supports the development of text mining systems for taxonomic studies, species distribution research, and comparative genomics applications. The corpus addresses the challenge of recognizing both scientific names and common names of organisms across diverse biological texts. Current Model Performance - F1 Score: `0.95` - Precision: `0.94` - Recall: `0.96` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-SpeciesDetect-PubMed-335M | 0.9649 | 0.9582 | 0.9718 | 0.9967 | | 🥈 2 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9543 | 0.9422 | 0.9667 | 0.9956 | | 🥉 3 | OpenMed-NER-SpeciesDetect-BioMed-335M | 0.9539 | 0.9441 | 0.9638 | 0.9957 | | 4 | OpenMed-NER-SpeciesDetect-SuperClinical-434M | 0.9534 | 0.9369 | 0.9704 | 0.9959 | | 5 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9502 | 0.9317 | 0.9695 | 0.9951 | | 6 | OpenMed-NER-SpeciesDetect-MultiMed-335M | 0.9479 | 0.9286 | 0.9680 | 0.9955 | | 7 | OpenMed-NER-SpeciesDetect-MultiMed-568M | 0.9460 | 0.9312 | 0.9613 | 0.9957 | | 8 | OpenMed-NER-SpeciesDetect-SuperMedical-355M | 0.9433 | 0.9221 | 0.9655 | 0.9953 | | 9 | OpenMed-NER-SpeciesDetect-SuperClinical-141M | 0.9406 | 0.9290 | 0.9525 | 0.9950 | | 10 | OpenMed-NER-SpeciesDetect-ModernClinical-395M | 0.9385 | 0.9379 | 0.9392 | 0.9940 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: LINNAEUS - Description: Species Entity Recognition - Species and organism names Training Details - Base Model: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.96` - Precision: `0.95` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: e5-large-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-large-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.82` - Precision: `0.79` - Recall: `0.85` - Accuracy: `0.96` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: e5-small-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-small-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.88` - Precision: `0.87` - Recall: `0.89` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: bge-m3 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-m3 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.86` - Precision: `0.86` - Recall: `0.86` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: BioClinical-ModernBERT-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.86` - Precision: `0.86` - Recall: `0.86` - Accuracy: `0.95` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.95` - Precision: `0.94` - Recall: `0.95` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: xlm-roberta-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.99` - Precision: `1.00` - Recall: `0.99` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: xlm-roberta-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Anatomical Entity Recognition - Anatomical structures and body parts [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for anatomical entity recognition - anatomical structures and body parts. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated ANATOMY dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Anatomy corpus focuses on anatomical entity recognition for medical terminology and healthcare applications. The Anatomy corpus is a specialized biomedical NER dataset designed for recognizing anatomical entities and medical terminology in clinical and biomedical texts. This corpus contains annotations for anatomical structures, body parts, organs, and physiological systems mentioned in medical literature. It is essential for developing clinical NLP systems, medical education tools, and healthcare informatics applications where accurate anatomical entity identification is crucial. The dataset supports the development of automated systems for medical coding, clinical decision support, and anatomical knowledge extraction from medical records and literature. It serves as a valuable resource for training NER models used in medical imaging, surgical planning, and clinical documentation. Current Model Performance - F1 Score: `0.86` - Precision: `0.86` - Recall: `0.86` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-AnatomyDetect-ElectraMed-560M | 0.9063 | 0.9083 | 0.9044 | 0.9825 | | 🥈 2 | OpenMed-NER-AnatomyDetect-PubMed-335M | 0.9063 | 0.8995 | 0.9131 | 0.9851 | | 🥉 3 | OpenMed-NER-AnatomyDetect-SuperClinical-434M | 0.9024 | 0.9040 | 0.9008 | 0.9836 | | 4 | OpenMed-NER-AnatomyDetect-ElectraMed-335M | 0.9020 | 0.9024 | 0.9016 | 0.9787 | | 5 | OpenMed-NER-AnatomyDetect-MultiMed-568M | 0.9012 | 0.8977 | 0.9048 | 0.9812 | | 6 | OpenMed-NER-AnatomyDetect-PubMed-109M | 0.9004 | 0.8941 | 0.9067 | 0.9844 | | 7 | OpenMed-NER-AnatomyDetect-SuperMedical-355M | 0.9002 | 0.8974 | 0.9029 | 0.9815 | | 8 | OpenMed-NER-AnatomyDetect-BigMed-560M | 0.8980 | 0.9007 | 0.8954 | 0.9814 | | 9 | OpenMed-NER-AnatomyDetect-BioMed-335M | 0.8961 | 0.8941 | 0.8982 | 0.9830 | | 10 | OpenMed-NER-AnatomyDetect-BioClinical-108M | 0.8961 | 0.8960 | 0.8962 | 0.9768 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: ANATOMY - Description: Anatomical Entity Recognition - Anatomical structures and body parts Training Details - Base Model: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.99` - Precision: `0.98` - Recall: `0.99` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: bge-large-en-v1.5 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-large-en-v1.5 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.87` - Precision: `0.86` - Recall: `0.88` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for clinical entity recognition - clinical entities related to chronic lymphocytic leukemia. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated CLL dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: CLL corpus is specialized for chronic lymphocytic leukemia entity recognition in hematology and cancer research. The CLL (Chronic Lymphocytic Leukemia) corpus is a domain-specific biomedical NER dataset focused on entities related to chronic lymphocytic leukemia, a type of blood cancer. This specialized corpus contains annotations for CLL-specific terminology, biomarkers, treatment entities, and clinical concepts relevant to hematology and oncology research. The dataset is designed to support the development of clinical NLP systems for leukemia research, hematological disorder analysis, and cancer informatics applications. It is particularly valuable for identifying disease-specific entities, therapeutic interventions, and prognostic factors mentioned in CLL research literature. The corpus serves as a benchmark for evaluating NER models in specialized medical domains and clinical research. Current Model Performance - F1 Score: `0.84` - Precision: `0.98` - Recall: `0.74` - Accuracy: `0.94` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-BloodCancerDetect-ElectraMed-560M | 0.9575 | 0.9264 | 0.9907 | 0.9843 | | 🥈 2 | OpenMed-NER-BloodCancerDetect-SuperClinical-434M | 0.8902 | 0.8652 | 0.9167 | 0.9701 | | 🥉 3 | OpenMed-NER-BloodCancerDetect-TinyMed-82M | 0.8793 | 0.7904 | 0.9908 | 0.9449 | | 4 | OpenMed-NER-BloodCancerDetect-TinyMed-135M | 0.8792 | 0.8750 | 0.8835 | 0.9668 | | 5 | OpenMed-NER-BloodCancerDetect-TinyMed-65M | 0.8547 | 0.7812 | 0.9434 | 0.9686 | | 6 | OpenMed-NER-BloodCancerDetect-SuperMedical-125M | 0.8488 | 1.0000 | 0.7373 | 0.9274 | | 7 | OpenMed-NER-BloodCancerDetect-SnowMed-568M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 8 | OpenMed-NER-BloodCancerDetect-BigMed-278M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 9 | OpenMed-NER-BloodCancerDetect-SuperMedical-355M | 0.8421 | 0.9816 | 0.7373 | 0.9248 | | 10 | OpenMed-NER-BloodCancerDetect-ElectraMed-335M | 0.8364 | 0.7302 | 0.9787 | 0.9581 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: CLL - Description: Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.99` - Precision: `1.00` - Recall: `0.99` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: roberta-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: roberta-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.91` - Precision: `0.90` - Recall: `0.91` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: BioClinicalBERT - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinicalBERT - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.94` - Precision: `0.93` - Recall: `0.94` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: xlm-roberta-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species names from the Species-800 dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species names from the species-800 dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated SPECIES800 dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Species800 is a corpus for species recognition and taxonomy classification in biomedical texts. The Species800 corpus is a manually annotated dataset designed for species recognition and taxonomic classification in biomedical literature. This corpus contains 800 abstracts with comprehensive annotations for organism mentions, supporting biodiversity informatics and biological taxonomy research. The dataset includes both scientific names and common names of species, making it valuable for developing NER systems that can handle the complexity of biological nomenclature. It serves as a benchmark for evaluating species identification models used in ecological studies, conservation biology, and systematic biology research. The corpus is particularly useful for text mining applications in biodiversity databases and biological literature analysis. Current Model Performance - F1 Score: `0.84` - Precision: `0.83` - Recall: `0.86` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OrganismDetect-BioMed-335M | 0.8639 | 0.8557 | 0.8722 | 0.9715 | | 🥈 2 | OpenMed-NER-OrganismDetect-PubMed-335M | 0.8550 | 0.8370 | 0.8737 | 0.9698 | | 🥉 3 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8458 | 0.8287 | 0.8637 | 0.9690 | | 4 | OpenMed-NER-OrganismDetect-MultiMed-335M | 0.8441 | 0.8352 | 0.8532 | 0.9670 | | 5 | OpenMed-NER-OrganismDetect-SuperClinical-434M | 0.8435 | 0.8291 | 0.8585 | 0.9670 | | 6 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8349 | 0.8082 | 0.8634 | 0.9685 | | 7 | OpenMed-NER-OrganismDetect-MultiMed-568M | 0.8313 | 0.8053 | 0.8592 | 0.9703 | | 8 | OpenMed-NER-OrganismDetect-ElectraMed-335M | 0.8288 | 0.8176 | 0.8404 | 0.9631 | | 9 | OpenMed-NER-OrganismDetect-BioPatient-108M | 0.8154 | 0.8140 | 0.8169 | 0.9591 | | 10 | OpenMed-NER-OrganismDetect-ElectraMed-33M | 0.8121 | 0.7772 | 0.8503 | 0.9600 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: SPECIES800 - Description: Species Entity Recognition - Species names from the Species-800 dataset Training Details - Base Model: deberta-v3-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species names from the Species-800 dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species names from the species-800 dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated SPECIES800 dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Species800 is a corpus for species recognition and taxonomy classification in biomedical texts. The Species800 corpus is a manually annotated dataset designed for species recognition and taxonomic classification in biomedical literature. This corpus contains 800 abstracts with comprehensive annotations for organism mentions, supporting biodiversity informatics and biological taxonomy research. The dataset includes both scientific names and common names of species, making it valuable for developing NER systems that can handle the complexity of biological nomenclature. It serves as a benchmark for evaluating species identification models used in ecological studies, conservation biology, and systematic biology research. The corpus is particularly useful for text mining applications in biodiversity databases and biological literature analysis. Current Model Performance - F1 Score: `0.77` - Precision: `0.73` - Recall: `0.83` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OrganismDetect-BioMed-335M | 0.8639 | 0.8557 | 0.8722 | 0.9715 | | 🥈 2 | OpenMed-NER-OrganismDetect-PubMed-335M | 0.8550 | 0.8370 | 0.8737 | 0.9698 | | 🥉 3 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8458 | 0.8287 | 0.8637 | 0.9690 | | 4 | OpenMed-NER-OrganismDetect-MultiMed-335M | 0.8441 | 0.8352 | 0.8532 | 0.9670 | | 5 | OpenMed-NER-OrganismDetect-SuperClinical-434M | 0.8435 | 0.8291 | 0.8585 | 0.9670 | | 6 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8349 | 0.8082 | 0.8634 | 0.9685 | | 7 | OpenMed-NER-OrganismDetect-MultiMed-568M | 0.8313 | 0.8053 | 0.8592 | 0.9703 | | 8 | OpenMed-NER-OrganismDetect-ElectraMed-335M | 0.8288 | 0.8176 | 0.8404 | 0.9631 | | 9 | OpenMed-NER-OrganismDetect-BioPatient-108M | 0.8154 | 0.8140 | 0.8169 | 0.9591 | | 10 | OpenMed-NER-OrganismDetect-ElectraMed-33M | 0.8121 | 0.7772 | 0.8503 | 0.9600 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: SPECIES800 - Description: Species Entity Recognition - Species names from the Species-800 dataset Training Details - Base Model: roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.88` - Precision: `0.86` - Recall: `0.89` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: deberta-v3-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for clinical entity recognition - clinical entities related to chronic lymphocytic leukemia. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated CLL dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: CLL corpus is specialized for chronic lymphocytic leukemia entity recognition in hematology and cancer research. The CLL (Chronic Lymphocytic Leukemia) corpus is a domain-specific biomedical NER dataset focused on entities related to chronic lymphocytic leukemia, a type of blood cancer. This specialized corpus contains annotations for CLL-specific terminology, biomarkers, treatment entities, and clinical concepts relevant to hematology and oncology research. The dataset is designed to support the development of clinical NLP systems for leukemia research, hematological disorder analysis, and cancer informatics applications. It is particularly valuable for identifying disease-specific entities, therapeutic interventions, and prognostic factors mentioned in CLL research literature. The corpus serves as a benchmark for evaluating NER models in specialized medical domains and clinical research. Current Model Performance - F1 Score: `0.76` - Precision: `0.71` - Recall: `0.83` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-BloodCancerDetect-ElectraMed-560M | 0.9575 | 0.9264 | 0.9907 | 0.9843 | | 🥈 2 | OpenMed-NER-BloodCancerDetect-SuperClinical-434M | 0.8902 | 0.8652 | 0.9167 | 0.9701 | | 🥉 3 | OpenMed-NER-BloodCancerDetect-TinyMed-82M | 0.8793 | 0.7904 | 0.9908 | 0.9449 | | 4 | OpenMed-NER-BloodCancerDetect-TinyMed-135M | 0.8792 | 0.8750 | 0.8835 | 0.9668 | | 5 | OpenMed-NER-BloodCancerDetect-TinyMed-65M | 0.8547 | 0.7812 | 0.9434 | 0.9686 | | 6 | OpenMed-NER-BloodCancerDetect-SuperMedical-125M | 0.8488 | 1.0000 | 0.7373 | 0.9274 | | 7 | OpenMed-NER-BloodCancerDetect-SnowMed-568M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 8 | OpenMed-NER-BloodCancerDetect-BigMed-278M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 9 | OpenMed-NER-BloodCancerDetect-SuperMedical-355M | 0.8421 | 0.9816 | 0.7373 | 0.9248 | | 10 | OpenMed-NER-BloodCancerDetect-ElectraMed-335M | 0.8364 | 0.7302 | 0.9787 | 0.9581 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: CLL - Description: Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia Training Details - Base Model: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.80` - Precision: `0.77` - Recall: `0.84` - Accuracy: `0.94` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: e5-small-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-small-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.96` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: xlm-roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.96` - Precision: `0.95` - Recall: `0.97` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: BiomedNLP-BiomedBERT-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene Entity Recognition - Gene-related entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene entity recognition - gene-related entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated GELLUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Gellus corpus targets gene recognition and genetics entities for genomics and molecular biology applications. The Gellus corpus is a biomedical NER dataset specifically designed for gene recognition and genetics entity extraction in molecular biology literature. This corpus contains comprehensive annotations for gene names, genetic variants, and genomics-related entities that are essential for genetic research and genomics applications. The dataset supports the development of automated systems for gene mention identification, genetic association studies, and genomics text mining. It is particularly valuable for identifying genes involved in hereditary diseases, genetic disorders, and molecular genetics research. The corpus serves as a benchmark for evaluating NER models used in genetics research, personalized medicine, and genomics informatics, contributing to advances in precision medicine and genetic counseling applications. Current Model Performance - F1 Score: `0.99` - Precision: `0.99` - Recall: `0.99` - Accuracy: `1.00` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomicDetect-SnowMed-568M | 0.9976 | 0.9977 | 0.9975 | 0.9989 | | 🥈 2 | OpenMed-NER-GenomicDetect-SuperMedical-355M | 0.9970 | 0.9960 | 0.9981 | 0.9986 | | 🥉 3 | OpenMed-NER-GenomicDetect-BigMed-560M | 0.9968 | 0.9967 | 0.9969 | 0.9986 | | 4 | OpenMed-NER-GenomicDetect-MultiMed-568M | 0.9967 | 0.9974 | 0.9960 | 0.9985 | | 5 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9964 | 0.9957 | 0.9970 | 0.9992 | | 6 | OpenMed-NER-GenomicDetect-PubMed-335M | 0.9963 | 0.9961 | 0.9965 | 0.9991 | | 7 | OpenMed-NER-GenomicDetect-PubMed-109M | 0.9951 | 0.9948 | 0.9953 | 0.9991 | | 8 | OpenMed-NER-GenomicDetect-BioMed-109M | 0.9941 | 0.9934 | 0.9949 | 0.9988 | | 9 | OpenMed-NER-GenomicDetect-TinyMed-82M | 0.9940 | 0.9997 | 0.9884 | 0.9961 | | 10 | OpenMed-NER-GenomicDetect-SuperMedical-125M | 0.9934 | 0.9999 | 0.9870 | 0.9958 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: GELLUS - Description: Gene Entity Recognition - Gene-related entities Training Details - Base Model: deberta-v3-small - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-small - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.94` - Precision: `0.94` - Recall: `0.95` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.00` - Precision: `0.01` - Recall: `0.00` - Accuracy: `0.63` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: e5-small-v2 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: e5-small-v2 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.93` - Precision: `0.92` - Recall: `0.94` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: deberta-v3-small - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-small - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.80` - Precision: `0.76` - Recall: `0.85` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-large-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRCHEM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Chem focuses on chemical entity recognition from the BioCreative V Chemical-Disease Relation extraction task. The BC5CDR-Chem corpus is part of the BioCreative V Chemical-Disease Relation (CDR) extraction challenge, specifically targeting chemical entity recognition in biomedical texts. This dataset contains 1,500 PubMed abstracts with 4,409 annotated chemical entities, designed to support automated drug discovery and pharmacovigilance applications. The corpus emphasizes chemical compounds, drugs, and therapeutic substances that are relevant for understanding chemical-disease relationships. It serves as a critical resource for developing NER systems that can identify chemical entities for downstream tasks like adverse drug reaction detection and drug repurposing research. Current Model Performance - F1 Score: `0.95` - Precision: `0.95` - Recall: `0.96` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PharmaDetect-SuperClinical-434M | 0.9614 | 0.9520 | 0.9710 | 0.9892 | | 🥈 2 | OpenMed-NER-PharmaDetect-MultiMed-335M | 0.9610 | 0.9585 | 0.9634 | 0.9871 | | 🥉 3 | OpenMed-NER-PharmaDetect-ElectraMed-335M | 0.9594 | 0.9539 | 0.9649 | 0.9863 | | 4 | OpenMed-NER-PharmaDetect-PubMed-335M | 0.9587 | 0.9521 | 0.9654 | 0.9902 | | 5 | OpenMed-NER-PharmaDetect-SuperMedical-355M | 0.9585 | 0.9520 | 0.9651 | 0.9881 | | 6 | OpenMed-NER-PharmaDetect-BioPatient-108M | 0.9583 | 0.9511 | 0.9656 | 0.9857 | | 7 | OpenMed-NER-PharmaDetect-ElectraMed-560M | 0.9562 | 0.9483 | 0.9642 | 0.9888 | | 8 | OpenMed-NER-PharmaDetect-BioClinical-108M | 0.9560 | 0.9504 | 0.9617 | 0.9849 | | 9 | OpenMed-NER-PharmaDetect-PubMed-109M | 0.9555 | 0.9417 | 0.9697 | 0.9889 | | 10 | OpenMed-NER-PharmaDetect-SuperMedical-125M | 0.9550 | 0.9442 | 0.9662 | 0.9871 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRCHEM - Description: Chemical Entity Recognition - Chemical entities from the BC5CDR dataset Training Details - Base Model: BioClinical-ModernBERT-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species and organism names [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species and organism names. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated LINNAEUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Linnaeus corpus is designed for species name identification and taxonomic entity recognition in biomedical literature. The Linnaeus corpus is a specialized biomedical NER dataset focused on species name identification and organism recognition in scientific literature. Named after Carl Linnaeus who established modern taxonomic nomenclature, this corpus contains annotations for species mentions that are normalized to NCBI Taxonomy identifiers. The dataset is crucial for biodiversity informatics, ecological research, and biological literature mining where accurate organism identification is essential. It supports the development of text mining systems for taxonomic studies, species distribution research, and comparative genomics applications. The corpus addresses the challenge of recognizing both scientific names and common names of organisms across diverse biological texts. Current Model Performance - F1 Score: `0.90` - Precision: `0.93` - Recall: `0.87` - Accuracy: `0.99` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-SpeciesDetect-PubMed-335M | 0.9649 | 0.9582 | 0.9718 | 0.9967 | | 🥈 2 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9543 | 0.9422 | 0.9667 | 0.9956 | | 🥉 3 | OpenMed-NER-SpeciesDetect-BioMed-335M | 0.9539 | 0.9441 | 0.9638 | 0.9957 | | 4 | OpenMed-NER-SpeciesDetect-SuperClinical-434M | 0.9534 | 0.9369 | 0.9704 | 0.9959 | | 5 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9502 | 0.9317 | 0.9695 | 0.9951 | | 6 | OpenMed-NER-SpeciesDetect-MultiMed-335M | 0.9479 | 0.9286 | 0.9680 | 0.9955 | | 7 | OpenMed-NER-SpeciesDetect-MultiMed-568M | 0.9460 | 0.9312 | 0.9613 | 0.9957 | | 8 | OpenMed-NER-SpeciesDetect-SuperMedical-355M | 0.9433 | 0.9221 | 0.9655 | 0.9953 | | 9 | OpenMed-NER-SpeciesDetect-SuperClinical-141M | 0.9406 | 0.9290 | 0.9525 | 0.9950 | | 10 | OpenMed-NER-SpeciesDetect-ModernClinical-395M | 0.9385 | 0.9379 | 0.9392 | 0.9940 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: LINNAEUS - Description: Species Entity Recognition - Species and organism names Training Details - Base Model: ModernBERT-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: ModernBERT-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.79` - Precision: `0.75` - Recall: `0.84` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.93` - Precision: `0.92` - Recall: `0.94` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedELECTRA-base-uncased-abstract - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - identifies chemical compounds and substances in biomedical literature. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC4CHEMD dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC4CHEMD is a biomedical NER corpus for chemical entity recognition from the BioCreative IV challenge. The BC4CHEMD (BioCreative IV Chemical Entity Mention) corpus is a manually annotated dataset designed for chemical entity recognition in biomedical literature. Created for the BioCreative IV challenge, this corpus contains abstracts from PubMed with chemical entities annotated according to Chemical Entities of Biological Interest (ChEBI) guidelines. The dataset is specifically designed to advance automated chemical name recognition systems for drug discovery, pharmacology, and chemical biology applications. It serves as a benchmark for evaluating named entity recognition models in identifying chemical compounds, drugs, and other chemical substances mentioned in scientific literature. Current Model Performance - F1 Score: `0.94` - Precision: `0.94` - Recall: `0.94` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ChemicalDetect-PubMed-335M | 0.9540 | 0.9498 | 0.9582 | 0.9902 | | 🥈 2 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9490 | 0.9447 | 0.9534 | 0.9891 | | 🥉 3 | OpenMed-NER-ChemicalDetect-PubMed-109M | 0.9487 | 0.9418 | 0.9557 | 0.9892 | | 4 | OpenMed-NER-ChemicalDetect-SnowMed-568M | 0.9485 | 0.9469 | 0.9502 | 0.9891 | | 5 | OpenMed-NER-ChemicalDetect-ElectraMed-560M | 0.9480 | 0.9455 | 0.9505 | 0.9890 | | 6 | OpenMed-NER-ChemicalDetect-SuperClinical-434M | 0.9469 | 0.9427 | 0.9512 | 0.9881 | | 7 | OpenMed-NER-ChemicalDetect-SuperMedical-355M | 0.9462 | 0.9418 | 0.9507 | 0.9875 | | 8 | OpenMed-NER-ChemicalDetect-MultiMed-335M | 0.9460 | 0.9435 | 0.9485 | 0.9857 | | 9 | OpenMed-NER-ChemicalDetect-MultiMed-568M | 0.9459 | 0.9437 | 0.9481 | 0.9885 | | 10 | OpenMed-NER-ChemicalDetect-BigMed-560M | 0.9454 | 0.9376 | 0.9534 | 0.9888 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC4CHEMD - Description: Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature Training Details - Base Model: BioDischargeSummaryBERT - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioDischargeSummaryBERT - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the NCBI dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the ncbi dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated NCBIDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: NCBI Disease corpus is a comprehensive resource for disease name recognition and concept normalization. The NCBI Disease corpus is a gold-standard dataset containing 793 PubMed abstracts with 6,892 disease mentions mapped to 790 unique disease concepts from Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM). Developed by the National Center for Biotechnology Information, this corpus provides both mention-level and concept-level annotations for disease entity recognition and normalization. The dataset is extensively used for developing clinical NLP systems, medical diagnosis support tools, and biomedical text mining applications. It serves as a critical benchmark for evaluating disease name recognition systems in healthcare informatics and medical literature analysis. Current Model Performance - F1 Score: `0.87` - Precision: `0.85` - Recall: `0.89` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9110 | 0.8918 | 0.9310 | 0.9792 | | 🥈 2 | OpenMed-NER-PathologyDetect-PubMed-335M | 0.9086 | 0.8913 | 0.9266 | 0.9781 | | 🥉 3 | OpenMed-NER-PathologyDetect-BioMed-335M | 0.9052 | 0.8867 | 0.9244 | 0.9780 | | 4 | OpenMed-NER-PathologyDetect-SuperClinical-434M | 0.9035 | 0.8772 | 0.9314 | 0.9760 | | 5 | OpenMed-NER-PathologyDetect-PubMed-109M | 0.9022 | 0.8825 | 0.9227 | 0.9769 | | 6 | OpenMed-NER-PathologyDetect-ElectraMed-335M | 0.8977 | 0.8884 | 0.9073 | 0.9719 | | 7 | OpenMed-NER-PathologyDetect-ElectraMed-560M | 0.8950 | 0.8749 | 0.9161 | 0.9747 | | 8 | OpenMed-NER-PathologyDetect-MultiMed-335M | 0.8903 | 0.8749 | 0.9063 | 0.9692 | | 9 | OpenMed-NER-PathologyDetect-SnowMed-568M | 0.8903 | 0.8684 | 0.9133 | 0.9731 | | 10 | OpenMed-NER-PathologyDetect-SuperClinical-141M | 0.8894 | 0.8633 | 0.9172 | 0.9744 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: NCBIDISEASE - Description: Disease Entity Recognition - Disease entities from the NCBI dataset Training Details - Base Model: xlm-roberta-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.77` - Precision: `0.75` - Recall: `0.79` - Accuracy: `0.91` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: deberta-v3-small - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: deberta-v3-small - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.77` - Precision: `0.73` - Recall: `0.81` - Accuracy: `0.92` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: BioClinical-ModernBERT-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BioClinical-ModernBERT-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for clinical entity recognition - clinical entities related to chronic lymphocytic leukemia. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated CLL dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: CLL corpus is specialized for chronic lymphocytic leukemia entity recognition in hematology and cancer research. The CLL (Chronic Lymphocytic Leukemia) corpus is a domain-specific biomedical NER dataset focused on entities related to chronic lymphocytic leukemia, a type of blood cancer. This specialized corpus contains annotations for CLL-specific terminology, biomarkers, treatment entities, and clinical concepts relevant to hematology and oncology research. The dataset is designed to support the development of clinical NLP systems for leukemia research, hematological disorder analysis, and cancer informatics applications. It is particularly valuable for identifying disease-specific entities, therapeutic interventions, and prognostic factors mentioned in CLL research literature. The corpus serves as a benchmark for evaluating NER models in specialized medical domains and clinical research. Current Model Performance - F1 Score: `0.82` - Precision: `0.90` - Recall: `0.76` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-BloodCancerDetect-ElectraMed-560M | 0.9575 | 0.9264 | 0.9907 | 0.9843 | | 🥈 2 | OpenMed-NER-BloodCancerDetect-SuperClinical-434M | 0.8902 | 0.8652 | 0.9167 | 0.9701 | | 🥉 3 | OpenMed-NER-BloodCancerDetect-TinyMed-82M | 0.8793 | 0.7904 | 0.9908 | 0.9449 | | 4 | OpenMed-NER-BloodCancerDetect-TinyMed-135M | 0.8792 | 0.8750 | 0.8835 | 0.9668 | | 5 | OpenMed-NER-BloodCancerDetect-TinyMed-65M | 0.8547 | 0.7812 | 0.9434 | 0.9686 | | 6 | OpenMed-NER-BloodCancerDetect-SuperMedical-125M | 0.8488 | 1.0000 | 0.7373 | 0.9274 | | 7 | OpenMed-NER-BloodCancerDetect-SnowMed-568M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 8 | OpenMed-NER-BloodCancerDetect-BigMed-278M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 9 | OpenMed-NER-BloodCancerDetect-SuperMedical-355M | 0.8421 | 0.9816 | 0.7373 | 0.9248 | | 10 | OpenMed-NER-BloodCancerDetect-ElectraMed-335M | 0.8364 | 0.7302 | 0.9787 | 0.9581 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: CLL - Description: Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia Training Details - Base Model: bge-m3 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-m3 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species names from the Species-800 dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species names from the species-800 dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated SPECIES800 dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Species800 is a corpus for species recognition and taxonomy classification in biomedical texts. The Species800 corpus is a manually annotated dataset designed for species recognition and taxonomic classification in biomedical literature. This corpus contains 800 abstracts with comprehensive annotations for organism mentions, supporting biodiversity informatics and biological taxonomy research. The dataset includes both scientific names and common names of species, making it valuable for developing NER systems that can handle the complexity of biological nomenclature. It serves as a benchmark for evaluating species identification models used in ecological studies, conservation biology, and systematic biology research. The corpus is particularly useful for text mining applications in biodiversity databases and biological literature analysis. Current Model Performance - F1 Score: `0.79` - Precision: `0.76` - Recall: `0.82` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OrganismDetect-BioMed-335M | 0.8639 | 0.8557 | 0.8722 | 0.9715 | | 🥈 2 | OpenMed-NER-OrganismDetect-PubMed-335M | 0.8550 | 0.8370 | 0.8737 | 0.9698 | | 🥉 3 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8458 | 0.8287 | 0.8637 | 0.9690 | | 4 | OpenMed-NER-OrganismDetect-MultiMed-335M | 0.8441 | 0.8352 | 0.8532 | 0.9670 | | 5 | OpenMed-NER-OrganismDetect-SuperClinical-434M | 0.8435 | 0.8291 | 0.8585 | 0.9670 | | 6 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8349 | 0.8082 | 0.8634 | 0.9685 | | 7 | OpenMed-NER-OrganismDetect-MultiMed-568M | 0.8313 | 0.8053 | 0.8592 | 0.9703 | | 8 | OpenMed-NER-OrganismDetect-ElectraMed-335M | 0.8288 | 0.8176 | 0.8404 | 0.9631 | | 9 | OpenMed-NER-OrganismDetect-BioPatient-108M | 0.8154 | 0.8140 | 0.8169 | 0.9591 | | 10 | OpenMed-NER-OrganismDetect-ElectraMed-33M | 0.8121 | 0.7772 | 0.8503 | 0.9600 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: SPECIES800 - Description: Species Entity Recognition - Species names from the Species-800 dataset Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.81` - Precision: `0.77` - Recall: `0.86` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: xlm-roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.79` - Precision: `0.74` - Recall: `0.84` - Accuracy: `0.93` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Various biomedical entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - various biomedical entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated FSU dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-protein` - `B-proteincomplex` - `B-proteinenum` - `B-proteinfamiliyorgroup` - `B-proteinvariant` - `I-protein` - `I-proteincomplex` - `I-proteinenum` - `I-proteinfamiliyorgroup` - `I-proteinvariant` FSU corpus focuses on protein interactions and molecular biology entities for systems biology research. The FSU (Florida State University) corpus is a biomedical NER dataset designed for protein interaction recognition and molecular biology entity extraction. This corpus contains annotations for proteins, protein complexes, protein families, protein variants, and molecular interaction entities relevant to systems biology and biochemistry research. The dataset supports the development of text mining systems for protein-protein interaction extraction, molecular pathway analysis, and systems biology applications. It is particularly valuable for identifying protein entities involved in cellular processes, signal transduction pathways, and molecular mechanisms. The corpus serves as a benchmark for evaluating NER systems used in proteomics research, drug discovery, and molecular biology informatics. Current Model Performance - F1 Score: `0.96` - Precision: `0.96` - Recall: `0.96` - Accuracy: `0.98` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-ProteinDetect-SnowMed-568M | 0.9609 | 0.9576 | 0.9642 | 0.9803 | | 🥈 2 | OpenMed-NER-ProteinDetect-ElectraMed-560M | 0.9609 | 0.9581 | 0.9636 | 0.9802 | | 🥉 3 | OpenMed-NER-ProteinDetect-MultiMed-568M | 0.9579 | 0.9564 | 0.9595 | 0.9788 | | 4 | OpenMed-NER-ProteinDetect-BigMed-560M | 0.9549 | 0.9520 | 0.9578 | 0.9778 | | 5 | OpenMed-NER-ProteinDetect-SuperMedical-355M | 0.9547 | 0.9517 | 0.9576 | 0.9749 | | 6 | OpenMed-NER-ProteinDetect-EuroMed-212M | 0.9482 | 0.9482 | 0.9482 | 0.9770 | | 7 | OpenMed-NER-ProteinDetect-BigMed-278M | 0.9466 | 0.9434 | 0.9499 | 0.9738 | | 8 | OpenMed-NER-ProteinDetect-SuperMedical-125M | 0.9465 | 0.9423 | 0.9507 | 0.9714 | | 9 | OpenMed-NER-ProteinDetect-SuperClinical-434M | 0.9412 | 0.9351 | 0.9474 | 0.9802 | | 10 | OpenMed-NER-ProteinDetect-TinyMed-82M | 0.9398 | 0.9331 | 0.9467 | 0.9680 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: FSU - Description: Biomedical Entity Recognition - Various biomedical entities Training Details - Base Model: bge-m3 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-m3 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species and organism names [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species and organism names. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated LINNAEUS dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Linnaeus corpus is designed for species name identification and taxonomic entity recognition in biomedical literature. The Linnaeus corpus is a specialized biomedical NER dataset focused on species name identification and organism recognition in scientific literature. Named after Carl Linnaeus who established modern taxonomic nomenclature, this corpus contains annotations for species mentions that are normalized to NCBI Taxonomy identifiers. The dataset is crucial for biodiversity informatics, ecological research, and biological literature mining where accurate organism identification is essential. It supports the development of text mining systems for taxonomic studies, species distribution research, and comparative genomics applications. The corpus addresses the challenge of recognizing both scientific names and common names of organisms across diverse biological texts. Current Model Performance - F1 Score: `0.72` - Precision: `0.66` - Recall: `0.79` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-SpeciesDetect-PubMed-335M | 0.9649 | 0.9582 | 0.9718 | 0.9967 | | 🥈 2 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9543 | 0.9422 | 0.9667 | 0.9956 | | 🥉 3 | OpenMed-NER-SpeciesDetect-BioMed-335M | 0.9539 | 0.9441 | 0.9638 | 0.9957 | | 4 | OpenMed-NER-SpeciesDetect-SuperClinical-434M | 0.9534 | 0.9369 | 0.9704 | 0.9959 | | 5 | OpenMed-NER-SpeciesDetect-PubMed-109M | 0.9502 | 0.9317 | 0.9695 | 0.9951 | | 6 | OpenMed-NER-SpeciesDetect-MultiMed-335M | 0.9479 | 0.9286 | 0.9680 | 0.9955 | | 7 | OpenMed-NER-SpeciesDetect-MultiMed-568M | 0.9460 | 0.9312 | 0.9613 | 0.9957 | | 8 | OpenMed-NER-SpeciesDetect-SuperMedical-355M | 0.9433 | 0.9221 | 0.9655 | 0.9953 | | 9 | OpenMed-NER-SpeciesDetect-SuperClinical-141M | 0.9406 | 0.9290 | 0.9525 | 0.9950 | | 10 | OpenMed-NER-SpeciesDetect-ModernClinical-395M | 0.9385 | 0.9379 | 0.9392 | 0.9940 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: LINNAEUS - Description: Species Entity Recognition - Species and organism names Training Details - Base Model: xlm-roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.88` - Precision: `0.88` - Recall: `0.88` - Accuracy: `0.94` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: snowflake-arctic-embed-l-v2.0 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: snowflake-arctic-embed-l-v2.0 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.85` - Precision: `0.82` - Recall: `0.87` - Accuracy: `0.95` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: distilroberta-base - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: distilroberta-base - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Cancer Genetics - Cancer-related genetic entities [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for cancer genetics - cancer-related genetic entities. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BIONLP2013CG dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-Aminoacid` - `B-Anatomicalsystem` - `B-Cancer` - `B-Cell` - `B-Cellularcomponent` - `B-Developinganatomicalstructure` - `B-Geneorgeneproduct` - `B-Immaterialanatomicalentity` - `B-Multi-tissuestructure` - `B-Organ` - `B-Organism` - `B-Organismsubdivision` - `B-Organismsubstance` - `B-Pathologicalformation` - `B-Simplechemical` - `B-Tissue` - `I-Aminoacid` - `I-Anatomicalsystem` - `I-Cancer` - `I-Cell` - `I-Cellularcomponent` - `I-Developinganatomicalstructure` - `I-Geneorgeneproduct` - `I-Immaterialanatomicalentity` - `I-Multi-tissuestructure` - `I-Organ` - `I-Organism` - `I-Organismsubdivision` - `I-Organismsubstance` - `I-Pathologicalformation` - `I-Simplechemical` - `I-Tissue` BioNLP 2013 CG corpus targets cancer genetics entities for oncology research and cancer genomics. The BioNLP 2013 CG (Cancer Genetics) corpus is a specialized dataset focusing on cancer genetics entities and gene regulation in oncology research. This corpus contains annotations for genes, proteins, and molecular processes specifically related to cancer biology and tumor genetics. Developed for the BioNLP Shared Task 2013, it supports the development of text mining systems for cancer research, oncological studies, and precision medicine applications. The dataset is particularly valuable for identifying cancer-related biomarkers, tumor suppressor genes, oncogenes, and therapeutic targets mentioned in cancer research literature. It serves as a benchmark for evaluating NER systems used in cancer genomics, personalized medicine, and oncology informatics. Current Model Performance - F1 Score: `0.88` - Precision: `0.88` - Recall: `0.89` - Accuracy: `0.94` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OncologyDetect-SuperMedical-355M | 0.8990 | 0.8926 | 0.9056 | 0.9416 | | 🥈 2 | OpenMed-NER-OncologyDetect-ElectraMed-560M | 0.8841 | 0.8788 | 0.8895 | 0.9390 | | 🥉 3 | OpenMed-NER-OncologyDetect-SnowMed-568M | 0.8801 | 0.8774 | 0.8828 | 0.9366 | | 4 | OpenMed-NER-OncologyDetect-PubMed-335M | 0.8782 | 0.8834 | 0.8730 | 0.9539 | | 5 | OpenMed-NER-OncologyDetect-MultiMed-568M | 0.8766 | 0.8749 | 0.8784 | 0.9351 | | 6 | OpenMed-NER-OncologyDetect-SuperClinical-434M | 0.8684 | 0.8602 | 0.8768 | 0.9495 | | 7 | OpenMed-NER-OncologyDetect-BioMed-335M | 0.8660 | 0.8540 | 0.8783 | 0.9516 | | 8 | OpenMed-NER-OncologyDetect-PubMed-109M | 0.8606 | 0.8604 | 0.8608 | 0.9503 | | 9 | OpenMed-NER-OncologyDetect-BigMed-560M | 0.8556 | 0.8582 | 0.8530 | 0.9250 | | 10 | OpenMed-NER-OncologyDetect-ModernClinical-395M | 0.8471 | 0.8465 | 0.8476 | 0.9411 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BIONLP2013CG - Description: Cancer Genetics - Cancer-related genetic entities Training Details - Base Model: multilingual-e5-large-instruct - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: multilingual-e5-large-instruct - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Gene/Protein Entity Recognition - Gene and protein mentions [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for gene/protein entity recognition - gene and protein mentions. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC2GM dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC2GM corpus targets gene and protein mention recognition from the BioCreative II Gene Mention task. The BC2GM (BioCreative II Gene Mention) corpus is a foundational dataset for gene and protein name recognition in biomedical literature, created for the BioCreative II challenge. This corpus contains thousands of sentences from MEDLINE abstracts with manually annotated gene and protein mentions, serving as a critical benchmark for genomics and molecular biology NER systems. The dataset addresses the challenging task of identifying gene names, which often have complex nomenclature and ambiguous boundaries. It has been instrumental in advancing automated gene recognition systems used in functional genomics research, gene expression analysis, and molecular biology text mining. The corpus continues to be widely used for training and evaluating biomedical NER models. Current Model Performance - F1 Score: `0.88` - Precision: `0.87` - Recall: `0.88` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-GenomeDetect-SuperClinical-434M | 0.9010 | 0.8954 | 0.9066 | 0.9683 | | 🥈 2 | OpenMed-NER-GenomeDetect-PubMed-335M | 0.8963 | 0.8924 | 0.9002 | 0.9719 | | 🥉 3 | OpenMed-NER-GenomeDetect-BioMed-335M | 0.8943 | 0.8887 | 0.8999 | 0.9704 | | 4 | OpenMed-NER-GenomeDetect-MultiMed-335M | 0.8905 | 0.8870 | 0.8940 | 0.9631 | | 5 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8894 | 0.8850 | 0.8937 | 0.9706 | | 6 | OpenMed-NER-GenomeDetect-BioPatient-108M | 0.8865 | 0.8850 | 0.8881 | 0.9590 | | 7 | OpenMed-NER-GenomeDetect-SuperMedical-355M | 0.8852 | 0.8802 | 0.8902 | 0.9668 | | 8 | OpenMed-NER-GenomeDetect-BioClinical-108M | 0.8851 | 0.8767 | 0.8937 | 0.9582 | | 9 | OpenMed-NER-GenomeDetect-MultiMed-568M | 0.8834 | 0.8770 | 0.8898 | 0.9671 | | 10 | OpenMed-NER-GenomeDetect-PubMed-109M | 0.8833 | 0.8781 | 0.8886 | 0.9706 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC2GM - Description: Gene/Protein Entity Recognition - Gene and protein mentions Training Details - Base Model: xlm-roberta-large - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: xlm-roberta-large - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for biomedical entity recognition - proteins, dna, rna, cell lines, and cell types. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated JNLPBA dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: - `B-DNA` - `B-RNA` - `B-cellline` - `B-celltype` - `B-protein` - `I-DNA` - `I-RNA` - `I-cellline` - `I-celltype` - `I-protein` JNLPBA corpus focuses on biomedical named entity recognition for protein, DNA, RNA, cell line, and cell type entities. The JNLPBA (Joint Workshop on Natural Language Processing in Biomedicine and its Applications) corpus is a widely-used biomedical NER dataset derived from the GENIA corpus for the 2004 bio-entity recognition task. It contains annotations for five entity types: protein, DNA, RNA, cell line, and cell type, making it essential for molecular biology and genomics research applications. The corpus consists of MEDLINE abstracts annotated with biomedical entities relevant to gene and protein recognition tasks. It has been extensively used as a benchmark for evaluating biomedical NER systems and continues to be a standard evaluation dataset for developing machine learning models in computational biology and bioinformatics. Current Model Performance - F1 Score: `0.82` - Precision: `0.78` - Recall: `0.86` - Accuracy: `0.94` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DNADetect-SuperClinical-434M | 0.8188 | 0.7778 | 0.8643 | 0.9320 | | 🥈 2 | OpenMed-NER-DNADetect-SuperMedical-355M | 0.8177 | 0.7716 | 0.8697 | 0.9318 | | 🥉 3 | OpenMed-NER-DNADetect-MultiMed-568M | 0.8157 | 0.7758 | 0.8599 | 0.9354 | | 4 | OpenMed-NER-DNADetect-BigMed-560M | 0.8134 | 0.7723 | 0.8591 | 0.9346 | | 5 | OpenMed-NER-DNADetect-BioClinical-108M | 0.8071 | 0.7632 | 0.8562 | 0.9147 | | 6 | OpenMed-NER-DNADetect-MultiMed-335M | 0.8069 | 0.7642 | 0.8547 | 0.9185 | | 7 | OpenMed-NER-DNADetect-PubMed-335M | 0.8056 | 0.7611 | 0.8556 | 0.9344 | | 8 | OpenMed-NER-DNADetect-SuperClinical-184M | 0.8053 | 0.7548 | 0.8630 | 0.9259 | | 9 | OpenMed-NER-DNADetect-BioPatient-108M | 0.8052 | 0.7605 | 0.8555 | 0.9137 | | 10 | OpenMed-NER-DNADetect-SuperMedical-125M | 0.8044 | 0.7589 | 0.8557 | 0.9284 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: JNLPBA - Description: Biomedical Entity Recognition - Proteins, DNA, RNA, cell lines, and cell types Training Details - Base Model: bge-m3 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-m3 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Species Entity Recognition - Species names from the Species-800 dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for species entity recognition - species names from the species-800 dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated SPECIES800 dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: Species800 is a corpus for species recognition and taxonomy classification in biomedical texts. The Species800 corpus is a manually annotated dataset designed for species recognition and taxonomic classification in biomedical literature. This corpus contains 800 abstracts with comprehensive annotations for organism mentions, supporting biodiversity informatics and biological taxonomy research. The dataset includes both scientific names and common names of species, making it valuable for developing NER systems that can handle the complexity of biological nomenclature. It serves as a benchmark for evaluating species identification models used in ecological studies, conservation biology, and systematic biology research. The corpus is particularly useful for text mining applications in biodiversity databases and biological literature analysis. Current Model Performance - F1 Score: `0.83` - Precision: `0.81` - Recall: `0.86` - Accuracy: `0.97` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-OrganismDetect-BioMed-335M | 0.8639 | 0.8557 | 0.8722 | 0.9715 | | 🥈 2 | OpenMed-NER-OrganismDetect-PubMed-335M | 0.8550 | 0.8370 | 0.8737 | 0.9698 | | 🥉 3 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8458 | 0.8287 | 0.8637 | 0.9690 | | 4 | OpenMed-NER-OrganismDetect-MultiMed-335M | 0.8441 | 0.8352 | 0.8532 | 0.9670 | | 5 | OpenMed-NER-OrganismDetect-SuperClinical-434M | 0.8435 | 0.8291 | 0.8585 | 0.9670 | | 6 | OpenMed-NER-OrganismDetect-PubMed-109M | 0.8349 | 0.8082 | 0.8634 | 0.9685 | | 7 | OpenMed-NER-OrganismDetect-MultiMed-568M | 0.8313 | 0.8053 | 0.8592 | 0.9703 | | 8 | OpenMed-NER-OrganismDetect-ElectraMed-335M | 0.8288 | 0.8176 | 0.8404 | 0.9631 | | 9 | OpenMed-NER-OrganismDetect-BioPatient-108M | 0.8154 | 0.8140 | 0.8169 | 0.9591 | | 10 | OpenMed-NER-OrganismDetect-ElectraMed-33M | 0.8121 | 0.7772 | 0.8503 | 0.9600 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: SPECIES800 - Description: Species Entity Recognition - Species names from the Species-800 dataset Training Details - Base Model: bge-m3 - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: bge-m3 - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Disease Entity Recognition - Disease entities from the BC5CDR dataset [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for disease entity recognition - disease entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated BC5CDRDISEASE dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: BC5CDR-Disease targets disease entity recognition from the BioCreative V Chemical-Disease Relation extraction corpus. The BC5CDR-Disease corpus is the disease-focused component of the BioCreative V Chemical-Disease Relation (CDR) task, containing 1,500 PubMed abstracts with 5,818 annotated disease entities. This manually curated dataset is designed to advance automated disease name recognition for medical diagnosis, pathology research, and clinical decision support systems. The corpus includes annotations for various disease types, medical conditions, and pathological states mentioned in biomedical literature. It serves as a benchmark for evaluating NER models in clinical and biomedical applications where accurate disease entity identification is crucial for medical informatics and healthcare analytics. Current Model Performance - F1 Score: `0.88` - Precision: `0.86` - Recall: `0.89` - Accuracy: `0.97` 🏆 Comparative Performance on BC5CDRDISEASE Dataset | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-DiseaseDetect-SuperClinical-434M | 0.9118 | 0.9028 | 0.9211 | 0.9839 | | 🥈 2 | OpenMed-NER-DiseaseDetect-PubMed-335M | 0.9097 | 0.8932 | 0.9268 | 0.9849 | | 🥉 3 | OpenMed-NER-DiseaseDetect-MultiMed-335M | 0.9022 | 0.8890 | 0.9159 | 0.9758 | | 4 | OpenMed-NER-DiseaseDetect-BioMed-335M | 0.9005 | 0.8887 | 0.9126 | 0.9838 | | 5 | OpenMed-NER-DiseaseDetect-BioClinical-108M | 0.8999 | 0.8862 | 0.9140 | 0.9723 | | 6 | OpenMed-NER-DiseaseDetect-PubMed-109M | 0.8994 | 0.8899 | 0.9091 | 0.9839 | | 7 | OpenMed-NER-DiseaseDetect-BioPatient-108M | 0.8991 | 0.8864 | 0.9121 | 0.9721 | | 8 | OpenMed-NER-DiseaseDetect-SuperClinical-184M | 0.8943 | 0.8687 | 0.9214 | 0.9812 | | 9 | OpenMed-NER-DiseaseDetect-SuperClinical-141M | 0.8921 | 0.8686 | 0.9170 | 0.9809 | | 10 | OpenMed-NER-DiseaseDetect-MultiMed-568M | 0.8909 | 0.8803 | 0.9017 | 0.9776 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: BC5CDRDISEASE - Description: Disease Entity Recognition - Disease entities from the BC5CDR dataset Training Details - Base Model: distilbert-base-cased - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: distilbert-base-cased - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!

Specialized model for Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia [](https://opensource.org/licenses/Apache-2.0) []() []() [](https://huggingface.co/OpenMed) This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for clinical entity recognition - clinical entities related to chronic lymphocytic leukemia. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research applications. 🎯 Key Features - High Precision: Optimized for biomedical entity recognition - Domain-Specific: Trained on curated CLL dataset - Production-Ready: Validated on clinical benchmarks - Easy Integration: Compatible with Hugging Face Transformers ecosystem This model can identify and classify the following biomedical entities: CLL corpus is specialized for chronic lymphocytic leukemia entity recognition in hematology and cancer research. The CLL (Chronic Lymphocytic Leukemia) corpus is a domain-specific biomedical NER dataset focused on entities related to chronic lymphocytic leukemia, a type of blood cancer. This specialized corpus contains annotations for CLL-specific terminology, biomarkers, treatment entities, and clinical concepts relevant to hematology and oncology research. The dataset is designed to support the development of clinical NLP systems for leukemia research, hematological disorder analysis, and cancer informatics applications. It is particularly valuable for identifying disease-specific entities, therapeutic interventions, and prognostic factors mentioned in CLL research literature. The corpus serves as a benchmark for evaluating NER models in specialized medical domains and clinical research. Current Model Performance - F1 Score: `0.80` - Precision: `0.73` - Recall: `0.88` - Accuracy: `0.96` | Rank | Model | F1 Score | Precision | Recall | Accuracy | |------|-------|----------|-----------|--------|-----------| | 🥇 1 | OpenMed-NER-BloodCancerDetect-ElectraMed-560M | 0.9575 | 0.9264 | 0.9907 | 0.9843 | | 🥈 2 | OpenMed-NER-BloodCancerDetect-SuperClinical-434M | 0.8902 | 0.8652 | 0.9167 | 0.9701 | | 🥉 3 | OpenMed-NER-BloodCancerDetect-TinyMed-82M | 0.8793 | 0.7904 | 0.9908 | 0.9449 | | 4 | OpenMed-NER-BloodCancerDetect-TinyMed-135M | 0.8792 | 0.8750 | 0.8835 | 0.9668 | | 5 | OpenMed-NER-BloodCancerDetect-TinyMed-65M | 0.8547 | 0.7812 | 0.9434 | 0.9686 | | 6 | OpenMed-NER-BloodCancerDetect-SuperMedical-125M | 0.8488 | 1.0000 | 0.7373 | 0.9274 | | 7 | OpenMed-NER-BloodCancerDetect-SnowMed-568M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 8 | OpenMed-NER-BloodCancerDetect-BigMed-278M | 0.8443 | 0.9816 | 0.7407 | 0.9372 | | 9 | OpenMed-NER-BloodCancerDetect-SuperMedical-355M | 0.8421 | 0.9816 | 0.7373 | 0.9248 | | 10 | OpenMed-NER-BloodCancerDetect-ElectraMed-335M | 0.8364 | 0.7302 | 0.9787 | 0.9581 | Rankings based on F1-score performance across all models trained on this dataset. Figure: OpenMed (Open-Source) vs. Latest SOTA (Closed-Source) performance comparison across biomedical NER datasets. NOTE: The `aggregationstrategy` parameter defines how token predictions are grouped into entities. For a detailed explanation, please refer to the Hugging Face documentation. Here is a summary of the available strategies: - `none`: Returns raw token predictions without any aggregation. - `simple`: Groups adjacent tokens with the same entity type (e.g., `B-LOC` followed by `I-LOC`). - `first`: For word-based models, if tokens within a word have different entity tags, the tag of the first token is assigned to the entire word. - `average`: For word-based models, this strategy averages the scores of tokens within a word and applies the label with the highest resulting score. - `max`: For word-based models, the entity label from the token with the highest score within a word is assigned to the entire word. For efficient processing of large datasets, use proper batching with the `batchsize` parameter: Batch Size Guidelines: - CPU: Start with batchsize=1-4 - Single GPU: Try batchsize=8-32 depending on GPU memory - High-end GPU: Can handle batchsize=64 or higher - Monitor GPU utilization to find the optimal batch size for your hardware - Dataset: CLL - Description: Clinical Entity Recognition - Clinical entities related to Chronic Lymphocytic Leukemia Training Details - Base Model: distilbert-base-uncased - Training Framework: Hugging Face Transformers - Optimization: AdamW optimizer with learning rate scheduling - Validation: Cross-validation on held-out test set - Base Architecture: distilbert-base-uncased - Task: Token Classification (Named Entity Recognition) - Labels: Dataset-specific entity types - Input: Tokenized biomedical text - Output: BIO-tagged entity predictions This model is particularly useful for: - Clinical Text Mining: Extracting entities from medical records - Biomedical Research: Processing scientific literature - Drug Discovery: Identifying chemical compounds and drugs - Healthcare Analytics: Analyzing patient data and outcomes - Academic Research: Supporting biomedical NLP research Licensed under the Apache License 2.0. See LICENSE for details. We welcome contributions of all kinds! Whether you have ideas, feature requests, or want to join our mission to advance open-source Healthcare AI, we'd love to hear from you. Follow OpenMed Org on Hugging Face 🤗 and click "Watch" to stay updated on our latest releases and developments. If you use this model in your research or applications, please cite the following paper: Proper citation helps support and acknowledge my work. Thank you!