davidsbatista/Annotated-Semantic-Relationships-Datasets

Organizes 15+ datasets across three supervision paradigms: traditional closed-class extraction (7-53 relation types), open information extraction (unbounded relations), and distant supervision approaches. Spans biomedical, encyclopedic, and pharmaceutical domains with Portuguese and English corpora annotated for nominal and entity pair relationships. Provides standardized formats compatible with supervised NLP model training, ranging from 225 Medline abstracts to 10,717+ annotated examples.

704 stars. No commits in the last 6 months.