[derek73]

The first title batch of phase 2j (#752): wing gains a dated given reading, so it is a title only inside a title phrase wherever it stands.

What changes

Phase 2i's decision 10 already reads a lone Wing at the head of a name as the given name (Wing Lau, Mr Wing Lau), so what was left for wing is the two places decision 10 reads a modifier as listed: after a comma, and ending the name. A word that is both a modifier and a possible given name is read by the head's modifier test in every part (a title only with a head title after it in its run, of after it, or in the object of of or the):

input	1.4.0 – 2.6.0	now
Lau, Wing Kai	title Wing, given Kai	given Wing, middle Kai
Lau, Wing	title Wing	given Wing
John Smith Wing.	title Wing. (2.3.0–2.6.0; 1.4.0–2.2.0 family)	family Wing.
Lau, Wing Commander	title Wing Commander	unchanged
Wing Commander Lau	title Wing Commander	unchanged

Alone after a comma wing reads as dean has since phase 2i, including the cost decision 10 measured for role words: John Smith, Wing reverses to given Wing, family John Smith, as John Smith, Dean does. Accepted because the listing form is what writes the word there: of FEC's 22 names writing WING after a comma, the 14 opening the part with it are all FAMILY, WING …, and ctgov writes none. Elsewhere the two follow different tests (a twin grid read 1,854 of 8,820 texts differently), so wing is not a template for the contested titles. No listed phrase holds wing, so nothing is reported (Wing Lau stays silent).

Pinned as rows, all as master read them, none written by a corpus name:

• LAU, WING MR keeps title WING MR (the modifier test takes a salutation as its head), where LAU, DEAN MR gives given DEAN.
• Accepted in rules/H9: The Hon. Wing Lau keeps title The Hon. Wing (object of the), and John Smith Wing. Jr. keeps title Wing. (H2's shape reads the dotted suffix as a head; John Smith Wing. MD gives family Wing.).

Measured

Against master 12384fae, py3.11, every local corpus with --base; counts and token shapes only.

• fec-individuals: 14 moved, all fixed: AA, WING 7, AA, WING AA… 6, AA, WING A. MR. 1.
• Committed corpora: 1 move, H9's former Accepted example Lau, Wing, plus 6 new rule examples.
• No moves: CLDR/GND agreement, PSC (1,039,174 scored), ctgov-officials and tse-donors. PSC scores spoken order, where 2i already took wing's 25 fixes.
• Spelling pairs: jrc-names 0 now alike / 0 now apart of 218,822 persons; nepal-candidates 0 / 0 of 126,981 pairs. No spelling writes Wing after a comma.

Docs

• rules/H9:
  • The modifier bullet now covers a modifier that may be a given name.
  • New Precedence bullet: the modifier test, not a contested word's style partner, decides a word in both sets (Hon. Wing Lau keeps given Wing).
  • The Accepted example moves to Hudson, Rock, and two Accepted costs are added (above).
  • Because carries the reason.
  • Signal line widened from "the leading run" to "its run of titles".
  • The rule budget is re-recorded: H9 grows (Then 245 → 265 words, Accepted 5 → 6 bullets, 162 → 211 words): the new clause and its costs.
• decisions/H9: dated entry with SUPERSEDES for the old Lau, Wing cost, and the Standing line amended.
• decisions/vocabulary-model: Standing line and a log pointer.
• Release log: the existing "lone title qualifier" bullet gains the Wing exception.
• Stale prose fixed:
  • vocab README's modifier kind;
  • the RFC's decision-10 line;
  • the title_readings, Lexicon.title_modifiers and config.titles.TITLE_MODIFIERS docstrings, which said every modifier is in TITLES.
• Counts recorded on the word: SSA 188, Wikidata 17 given; Census 12,614, Wikidata 218 family; nam_dict band 1 in China.

Checks

uv run pytest (26,335 passed), -n0 -m serial, --python 3.15, mypy, ruff, sphinx-build -b doctest. The design-docs review (two rounds) ran a 26,460-parse shape grid; its findings (undersold reach, the reversal, the both-sets precedence, stale prose, then an overclaimed "reads as dean" and the dotted-suffix spelling) are fixed in this commit.

Refs #752, #853

🤖 Generated with Claude Code