SMI: Ontology-Grounded Semantic IDs for Hierarchy-Aware Medical Concept Representation
We introduce Semantic Medical Identifier (SMI), a training-free framework for constructing ontology-grounded Semantic IDs and hierarchy-aware medical concept representations.
ProofPaper ↗
Key points
- Biomedical language models (LMs) provide strong representations for clinical tasks, but they often fail to capture the hierarchical structure among medical concepts.
- For each diagnosis, procedure, or medication concept, SMI defines its Semantic ID as the coarse-to-fine path through an expert-defined medical ontology.
- SMI then aggregates frozen biomedical LM embeddings across the corresponding ontology levels to form hierarchy-aware medical concept representations that preserve coarse-to-fine clinical semantics.
- Experiments show that SMI captures meaningful hierarchical structure and improves clinical relation prediction.
Sources (1)
- [1]SMI: Ontology-Grounded Semantic IDs for Hierarchy-Aware Medical Concept RepresentationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 04:30 PM
We introduce Semantic Medical Identifier (SMI), a training-free framework for constructing ontology-grounded Semantic IDs and hierarchy-aware medical concept representations.
Biomedical language models (LMs) provide strong representations for clinical tasks, but they often fail to capture the hierarchical structure among medical concepts.
Extractive summary: sentences quoted from the sources.