DatasetWhich societies it draws from
IDSA, AAP, AUA, SCCM, KDIGO and ASCO are among those in "other."
How it was built
The flow is the same as the Japanese version. Enumerate titles for 40 societies on PubMed and cluster them into revision series. Gather full text through open access and publisher APIs, chunk it sentence-aware, and retrieve with Qwen3 embeddings. Gemma writes both editions' answers and takes the difference, at a retrieval depth of k=20. Each change is tagged with an axis marking what moved.
What's in a row
- type
- revision 4,745 / new 6,795 / none 6,071 / addition 4,770 / unrelated 1,219
- question
- a clinical question raised from the new edition
- new / old
- PMID · DOI · year · title · org · source URL · answer · source passage
- magnitude
- major / minor
- format
- Parquet · 121MB · English
- license
- CC BY-NC 4.0 (non-commercial · requires agreeing to terms)
License and a note
Unlike the Japanese version, this one is CC BY-NC 4.0 (non-commercial) and requires agreeing to the terms to use. The content is not a redistribution of the guideline text itself — it's differences and Q&A grounded in the retrieved passages. Check each row's source_url before any clinical use.