[derek73]

Classifies the 644 titles still generic in nameparser/vocab/und.toml into the ten title kinds #769 added, and moves each into the files of the languages whose names it occurs in. und.toml now holds no title (uv run python tools/vocab/query.py --kind title --file und --count prints 0), which is #750's done-when. One word keeps the generic kind, in en.toml: wm, the abbreviation of William, made a title for #68. No title kind describes it, and removing it would move a parse (the bare Wm Smith), so that is left to its own change.

No compiled set moves. tools/vocab/build.py --check reports _compiled.py current, and tools/differential/snapshot.py matches the snapshot. The suite, mypy and ruff are green.

What moved where

• en.toml takes most of them (530 of the 678 readings written): the bulk import's occupations, the US military ranks and their abbreviations, the offices, and the modifiers (vice, rear, petty, the ordinals, the domain words such as baseball and tax, and the court words from Judge-related titles not parsing #9's list of judges' titles).
• mul.toml: dr, prof, docent, doctor, professor.
• Noble ranks, native-script academic and religious titles go to their languages' files. Transliterations go to their source language, as in part 1: the Islamic titles to ar or fa, rabbi to he, rebbe to yi, maharajah/maharani/subedar to hi, tsar/tsarina to ru, effendi to tr.
• Four new files: bin (edohen, ekegbian), ff (lamido), mi (rangatira), yi (rebbe).
• Part 1's decisions applied: lord, lady, knight and chief are nobility, and deputy is an office.

Contested assignments (shown to Derek before opening, approved in chat 2026-10-19)

• In 3 or more files:
  • baron: en, fr, nl, de
  • sultan: ar, ms, tr
  • greve: sv, da, nb
  • д-р, проф, акад: ru, uk, bg
• Gender:
  • Noble ranks: the man's title of a pair is marked m, as king already is.
  • Descriptive nouns: only the -ess form is marked (actress, heiress).
  • sultan records none, because Ottoman usage gave it to imperial women too.
• academic as a form of address:
  • English: advocate, attorney, engineer, nurse.
  • German: ra, wp, pd/priv.-doz, dipl.-ing.
  • French: me.
  • Devanagari: पंडित/पं, matching part 1's academic refusal of the Latin pandit.
• Buddhist, Jain and Zoroastrian titles stay in en.toml as English loanwords.
• Also given names. Many moved words are given names (US SSA births: marquis 25,041, royal 24,394, deacon 11,570, baron 7,059, ...). This PR only classifies them; whether any should stop being a title is a separate change.

Comments became fields

• The Provide constants in non-Latin scripts (Cyrillic, Greek, Arabic, Hebrew) #269, Add Bengali honorifics to the default vocabulary #343 and Add the remaining Devanagari honorifics and the जी suffix #344 block comments are notes on one entry of each group.
• duke's keep-out comment is an excluded entry in en.toml, with its counts.
• und.toml's credential readings that shared an entry with a moved title sit in their own section, as Classify the particles in und.toml by language and kind #749's did.

The record is the new 2026-10-19 entry in docs/design/decisions/vocabulary-model.md. Review: the docs review found three small defects, fixed in the second commit and re-reviewed clean. The PR review (code reviewer and comment analyzer) found note, kind and doc defects, fixed in the third commit with Derek's calls on wm, doctor and पंडित. The review of that fix found three more (the wm parse claim, measured with the word listed rather than removed; two "all titles classified" statements; डा's remark in expands), fixed in the fourth; the fifth corrects part 1's recompute sentence, since --kind title counts the whole category and no query isolates the generic kind. Each fix commit was reviewed. The sixth records the wm lesson as a field note in docs/design/mechanisms.md's Verification shapes: a claim about what a vocabulary entry does is measured on the parse without it (docs review clean).

Closes #750.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.05%. Comparing base (732a271) to head (8051e3c).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #773   +/-   ##
=======================================
  Coverage   99.05%   99.05%           
=======================================
  Files          47       47           
  Lines        4642     4642           
=======================================
  Hits         4598     4598           
  Misses         44       44           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.