pruas/BENT-PubMedBERT-NER-Chemical

Named Entity Recognition (NER) model to recognize chemical entities.

Please cite our work:

@article{NILNKER2022,
  title = {NILINKER: Attention-based approach to NIL Entity Linking},
  journal = {Journal of Biomedical Informatics},
  volume = {132},
  pages = {104137},
  year = {2022},
  issn = {1532-0464},
  doi = {https://doi.org/10.1016/j.jbi.2022.104137},
  url = {https://www.sciencedirect.com/science/article/pii/S1532046422001526},
  author = {Pedro Ruas and Francisco M. Couto},
}

PubMedBERT fine-tuned on the following datasets:

• Chemdner patents CEMP corpus (train, dev, test sets)
• DDI corpus (train, dev, test sets): entity types "GROUP", "DRUG", "DRUG_N"
• GREC Corpus (train, dev, test sets): entity type "organic_compounds"
• MLEE (train, dev, test sets): entity type "Drug or compound"
• NLM-CHEM (train, dev, test sets)
• CHEMDNER (train, dev, test sets)
• Chebi Corpus (train, dev, test sets): entity types "Metabolite", "Chemical"
• PHAEDRA (train, dev, test sets): entity type "Pharmalogical_substance"
• Chemprot (train, dev, test sets)
• PGx Corpus (train, dev, test sets): entity type "Chemical"
• BioNLP11ID (train, dev, test sets): entity type "Chemical"
• BioNLP13CG (train, dev, test sets): entity type "Chemical"
• BC4CHEMD (train, dev, test sets)
• CRAFT corpus (train, dev, test sets): entity type "ChEBI"
• BC5CDR (train, dev, test sets): entity type "Chemical"