Automating Biomedical Knowledge Graph Construction For Context-Aware Scientific Inference
Yan Li, Yichun Feng, Wanquan Liu, Lu Zhou, Xiawei Du, Bi Zeng
bioRxiv (Cold Spring Harbor Laboratory)
Problems Identified (4)
Context-agnostic biomedical extraction: Existing biomedical extraction methods reduce dynamic, context-dependent interactions to binary associations that lose semantics and create contradictory evidence.
Complex contextual biomedical QA: Biomedical question answering includes complex queries that require fine-grained contextual information.
Context-agnostic biomedical extraction: Existing biomedical extraction methods reduce dynamic, context-dependent interactions to binary associations that lose semantics and create contradictory evidence.
Complex contextual biomedical QA: Biomedical question answering includes complex queries that require fine-grained contextual information.
Proposed Solutions (5)
AutoBioKG context-aware KG framework: AutoBioKG is an end-to-end framework for constructing context-aware biomedical knowledge graphs.
Composite triplet context encoding: The framework uses composite triplets to encode environmental conditions and entity attributes together with core relationships.
Self-evolving biomedical OpenIE: The framework is powered by a self-evolving open information extraction model trained on the curated BioOpenIE dataset.
AutoBioKG context-aware KG framework: AutoBioKG is an end-to-end framework for constructing context-aware biomedical knowledge graphs.
Composite triplet context encoding: The framework uses composite triplets to encode environmental conditions and entity attributes together with core relationships.
Results (3)
Zero-shot IE F1 improvement:
Biomedical QA improvement:
Scalable literature-to-knowledge solution:
Research Domain
Biomedical knowledge graph construction and biomedical information extraction