# Dictionary sources and licences

The derived `data/medical-words.json` dictionary is distributed under GNU GPL version 3. It is built from:

- Glutanimate, wordlist-medicalterms-en: https://github.com/glutanimate/wordlist-medicalterms-en
- Glutanimate, hunspell-en-med-glut: https://github.com/glutanimate/hunspell-en-med-glut
- Chris Barlow, medical-dictionary: https://github.com/cjdbarlow/medical-dictionary

Upstream credits include R. Robinson / e-MedTools (OpenMedSpel), Rajasekharan N. (MTH-Med-Spel-Chek / Raj&Co), Aristotelis P. / Glutanimate, the Hunspell Project, and Chrisjake. Consult the original headers and README files for the complete attribution.

Exact revisions and SHA-256 checksums are in `vendor/dictionaries/manifest.json`. Original files, copyright notices, licence texts, and READMEs are provided alongside the application in `vendor/dictionaries/`. The transformation source is `scripts/build-dictionaries.js`; run it with Node.js from the project root to reproduce the derived dictionary. The changes are lowercasing, apostrophe normalisation, token filtering, removal of counts/prose/affix flags/compound-only entries, and deduplication. Hunspell affix rules are preserved as source but are not expanded by this application. Non-token chemical expressions are excluded from the runtime word list.

LITFL's source includes general English as well as medical vocabulary; counts are dictionary entries, not a count of distinct medical concepts. The overlapping Glutanimate sources document drug coverage through April 2014. These datasets are not a current clinical terminology authority and may contain errors.

General English: `data/words.json` is derived from macOS `/usr/share/dict/words`, based on Webster's Second International. Its accompanying system README states the 1934 copyright has lapsed. This historical dictionary is separate from the GPL medical dataset.

The GPL dictionaries are supplied without warranty. See the accompanying full GPL texts for licence terms.

## Human genes

HGNC (HUGO Gene Nomenclature Committee, RRID:SCR_002827), https://www.genenames.org/. Data released under CC0-1.0: https://www.genenames.org/about/license/. The separate derived dataset `data/human-genes.json` retains CC0 status. Retrieval date, official source URL, checksums and transformation details are recorded in `vendor/hgnc/manifest.json`. Refresh with `python3 scripts/fetch-hgnc.py`. Only approved human records contribute symbols, historical/alias symbols, and gene-name vocabulary tokens. No gene identity, alias resolution or symbol correction is inferred.