[derek73]

The fourth title batch of phase 2j (#752): king gains a dated given reading (US SSA 37,890 births; Census rank 34 as a surname), so it is a title only inside a title phrase.

What changes

input	1.4.0 – 2.6.0	now
King Curtis (#348)	title King, given Curtis	given King, family Curtis
Mr King Smith (#853)	title Mr King	title Mr, given King
SMITH, KING	title KING	given KING
His Majesty King Charles	title His Majesty King	unchanged (a style is a partner)
King of England	title	unchanged
King Henry	title King, given Henry	given King, family Henry (the accepted cost)

No listed phrase holds king, so nothing is reported. The other sites read as for mahdi and buddha:

• before a family comma (King Smith, John → family King Smith, 1.4.0's reading);
• ending the name;
• alone after a full name and a comma;
• past a second comma.

The cost (rules/H9, Accepted; Derek, 2026-10-26)

A monarch written with the bare title reads it as the given name:

• King Henry, King Charles III and King John V.;
• King and Queen Smith → given King and Queen.

queen is in no batch, so Queen Elizabeth II keeps its title.

Where it falls:

• Committed corpora: three rows.
• jrc-names (news text): 20 persons open with King, about half monarchs or legendary kings and half people whose given name, stage name or nickname is King.
• PSC, FEC, ctgov and ukec: none.

pope replaces king as the worked example

king was the repository's example of a title that is also a surname. It appeared in rules H1, H3–H8, M2, S2 and S8, their decisions Standing lines, the H Background, AGENTS.md, docs/usage.rst, two code comments, case rows, five v1 tests and pipeline tests.

Each now uses pope. I checked 19 inputs against eight candidates (queen, bishop, pope, baron, judge, sheriff, mayor, emperor). Only pope and queen read exactly as king did on master; pope is the commoner surname. Dated Log entries keep King.

decisions/v1-xfail-triage's "king stays in TITLES" is superseded there, with a Log entry.

Measured

Against master 957509d1 (after #899 and #900), py3.11, every local corpus with --base; counts and token shapes only.

• PSC (1,039,174 scored): 58 records fixed, 0 broken, 5 changed.
• fec-individuals: 73 moved, all fixes.
  • 51 are AA, KING … given names.
  • 22 are KING AA, AA family parts that H6 had peeled.
• ctgov-officials: 12 moved, all fixes. Most are officials whose given name is King; one had its credential read as the given name.
• No moves: tse-donors, ukec-individuals, jrc-names/nepal pairs (0 / 0: both spellings of a person move together), CLDR/GND agreement.
• Committed corpora: 6 rows move: the three royal rows; King Curtis, fixed; and Dr King Jr and Dr. King, which keep their fields and lose H4's title-or-name. The rule examples rewritten from King to Pope leave the corpus (10) and join it (14), each Pope row reading as its King twin did on master.

Docs

• rules/H9: an example and an Accepted bullet for the royal cost. The rule budget is re-recorded: H9 Accepted grows from 8 to 9, and S2 grows by 8 words.
• decisions/H9: the dated entry with counts, cost, Derek's decision and the measurement.
• decisions/vocabulary-model and decisions/v1-xfail-triage: pointer and supersession entries.
• Release log: a new bullet.
• The king entry: given and family counts recorded on the word.
• Case rows: 6 new, 12 swapped to Pope (each classification re-checked against 1.4.0).

Checks

• uv run pytest and --python 3.15: 27,013 passed each.
• Also: -n0 -m serial, mypy, ruff, sphinx-build -b doctest, README doctest.
• A design-docs review in two rounds. Its findings are fixed here: the swap's misses in the H Background, usage.rst, an M2 Standing example, vocabulary-exclusions and a code comment; five tests left vacuous; two parity labels; one cross-version claim.

Refs #752, #348, #853

🤖 Generated with Claude Code