CIHRED: Complementary and Integrative Health Relation Extraction Dataset for Biomedical Text Mining

This dataset provides sentence-level annotations for Named Entity Recognition (NER) and Relation Extraction (RE) in the domain of Complementary and Integrative Health (CIH), sourced from PubMed abstracts. It is split into training, validation, and test sets using a document-level 60/20/20 split to ensure non-overlapping abstracts across splits.Each record contains the original sentence text, annotated entity mentions with character-level offsets and entity types (e.g., Mindbody_therapy, Disease, Chemical, Outcome_marker), and semantic relation triples in (subject, predicate, object) format with corresponding entity types and IDs.