Summary
Current words_cui.tsv is from April 2021. Download fresh 2025/2026 UMLS release using new UMLS account credentials. Rebuild umls_automata.bin.zst artifact.
Details
- Current automata has ~1.4M patterns from 2021 UMLS release
- New release will have updated concepts, retired CUIs, and new terms
- Validate pattern count and entity extraction quality against 18 evaluation cases
- Ensure no regression in safety gate behavior (e.g., Pembrolizumab/EGFR blocking)
Acceptance Criteria
Priority: P2
Summary
Current
words_cui.tsvis from April 2021. Download fresh 2025/2026 UMLS release using new UMLS account credentials. Rebuildumls_automata.bin.zstartifact.Details
Acceptance Criteria
umls_automata.bin.zstwith updated dataPriority: P2