Multi-Entity Biomedical NER with Distant Supervision
- Comparing Training Strategies for Schema Expansion toward Virus-Centric Drug Repurposing.
NERDistant SupervisionBiomedicalBERTDrug Repurposing
- Status
- Research
- Date
- Jun 2026
- Role
- Researcher and Developer
About
BC5CDR ships with two entity types: Chemical and Disease. For virus-centric drug repurposing the model also needs Virus and Gene. This project expands the entity schema by attaching distant-supervision silver labels to a scraped PubMed corpus, then compares several ways of mixing the original gold data with the noisier silver data so that a single biomedical NER model recognises all four entity types at once.