Vocabulary model and language packs (tracking) #747

Goal. Describe each vocabulary word by its readings — what the word can be in a name (a generation suffix, a credential, a family particle, a salutation…) and the properties of each reading — instead of base sets plus marker subsets. Then build language packs on top: choosing a language settles the words that are ambiguous only because the default serves every language.

The default keeps working for names of any language, as today. A pack prefers its language, settling clashes where that language has a vocabulary reading; a strict pack also drops other languages' name readings.

Where things are

• The model and every decision so far: decisions/vocabulary-model
• The data: nameparser/vocab/<language>.toml, compiled by tools/vocab/build.py (the parser never reads the TOML)
• A reference for editing the TOML and a design overview of the later phases: the docs sub-issue below

Status

Classification so far (master d3730550): 240 words have a language, almost all of them suffixes. 1,544 are still in und.toml under generic kinds: 801 titles, 605 credentials, 111 particles, and a few others. Recompute with uv run python tools/vocab/query.py --kind title --count, and likewise for other kinds.

Sub-issues track each piece. Related questions are linked under the phase they belong to.