[derek73]

Part of #1115, its second batch: the titles the mining found that no source counts as a name. The batch that decides which pack titles the default reads, the exclusions and the Diplom degrees come later.

What changes

Eight title readings, each with under 100 US Census 2010 bearers, under 5 US SSA births a year and at most 2 on Wikidata (tools/name_usage.py --boxes, 2026-11-06):

Word	File	Kind
engr, adv	en.toml	academic (beside engineer, advocate)
assist	en.toml	modifier (beside assistant)
ir, drs	mul.toml	academic: the Dutch and Indonesian degrees, Hong Kong's and Malaysia's registered engineer, the English plural of Dr
excm, excma	ca.toml	style
excmo, excma	es.toml	style

At the head of a name, the dotted spelling of each was already a title by rules/H2's shape. Listing reaches where the shape doesn't:

Engr Ahsen Ali                          given 'Engr', middle 'Ahsen'     →  title 'Engr', given 'Ahsen'
Prof dr ir Simon Peeters                title 'Prof dr', given 'ir'      →  title 'Prof dr ir', given 'Simon'
Esra Oz, Assist Prof                    given 'Assist', family 'Esra Oz' →  title 'Assist Prof', given 'Esra', family 'Oz'
BAXTER, JENNIFER DRS.                   middle 'DRS.'                    →  title 'DRS.'
Excm. Sr. Lloret Palol, Josep Antoni    family 'Excm. Lloret Palol', suffix 'Sr.'  →  title 'Excm. Sr.', family 'Lloret Palol'

Left out, borne as names:

• eng (US Census 10,682)
• er (ORCID family boxes 1,087)
• doc (US SSA 1,391; Doc Rivers)
• ven (Wikidata 448 given names)
• em (US Census 885)

Giving them a title reading beside a name reading (gated, rules/H9) was tried in a test lexicon and changes nothing at the head of a name. With abbreviation = true it adds only the dotted form ending a name, so they stay out. The unlisted Spanish and Czech titles (Licda., Gral., RSDr., plk.) wait for the pack-defaults batch.

Measured

uv run python tools/differential/config_diff.py diff --local all --committed --match '(?i)(^|[^\w])(engr|adv|ir|drs|assist|excm|excma|excmo)\b' '<the tree with the eight readings removed>' default, 2026-11-06 against e76cdad3. 246 of the 662 matching names move, 175 of them ClinicalTrials.gov officials written Name, Assist Prof. Every move was read by eye, and nearly all are a title leaving a name field. The costs, recorded in decisions/H6:

• four tokens that glue a title to a name word with a period take the word into the title, as Dr. already does (Engr.Nurunnabi → title 'Engr.Nurunnabi')
• one credential run splits (…, Reg Engr, MIMechE)
• eleven officials writing Res. Assist. read title 'Assist.', suffix 'Res.', as the listed Res. Asst. already did

The committed snapshot moves the two CLDR Excm./Excma. names, and the CLDR agreement record is re-recorded: both now read their title and family, closer to CLDR's fields. CLDR files Excm. as a credential.

Review

The design-docs review recomputed every count and the measurement (both exact), read all 246 moves and ran a grid of 1,008 parses. It found six problems in how the docs describe the moves, all fixed in the second commit:

• the release bullet's Assist limit, which was false outside the leading run
• the entry's "every move clean", which omitted the costs above
• the Res. Assist count (eleven, not two)
• one before-reading
• the comma mechanism (C8, not a suffix comma)
• ThDr., which is already listed

Test plan

• uv run pytest (39,128 passed) and uv run pytest -n0 -m serial
• uv run mypy, uv run ruff check, python -m doctest README.rst
• uv run sphinx-build -b doctest docs and the html build
• Case rows checked against an export of master; the release-log examples checked against the 1.4.0 and 2.8.2 wheels (the reviewer also ran 2.0.0, 2.3.0, 2.5.0 and 2.7.0)

🤖 Generated with Claude Code