data/README.md

# Bengali game dictionary

`bn-words.txt` is the normalized candidate pool generated from `BengaliWordList_40.txt` in
[MinhasKamal/BengaliDictionary](https://github.com/MinhasKamal/BengaliDictionary),
with the small hand-reviewed demo list merged in and entries normalized to NFC.

The upstream dataset is licensed under GNU GPL v3.0. This word list is candidate
lexical data for development; inclusion does not by itself mean a word is suitable
for a published puzzle. Published puzzles still require editorial review.

Confirmed OCR errors and invalid forms are removed by the explicit denylist in
`scripts/build-bengali-wordlist.mjs`. The source list is a candidate pool, not an
authoritative modern Bangla dictionary.

The game loads `bn-editorial-words.txt`, a conservative automated editorial view.
It excludes explicit source errors, known regional/colloquial forms, and common
case/plural suffix forms when their base word is also present. The full audit trail
is written to `bn-editorial-report.json`. Ambiguous words still require human review.