[derek73]

Summary

Follow-up to #927. The corpus it added has two consequences for the design docs, and this PR records both.

• docs/design/mechanisms.md#NAME-DATA-JOBS now covers orcid-researchers. It differs from the other local corpora in three ways:

  • its names are split by their bearers, not a filer, so a miss is a disagreement with the person;
  • it is capped at 20,000 records a country, so it shows each country's convention and never a frequency;
  • its country is where a holder works or lives.

  It is the SCORE to read for a language pack.

• docs/design/decisions/O6 has a dated log entry measuring Policy.two_family_names on the corpus by country, with a recompute recipe. PSC held only 116 Peruvian filers; ORCID holds 20,000 records for each of these countries.

  • The Spanish-speaking countries other than Argentina rise from 13.4–48.0% to 85.6–93.7% agreement.
  • It breaks Argentina (80.3% → 59.7%) and the Philippines (94.7% → 54.9%). The Philippines is a population PSC barely held: the holders file a middle name with their given names.
  • Portugal reads the other way from PSC's Portuguese filers.

• docs/customize.rst: the caution paragraph of the two_family_names section gains the Philippines and the Portuguese contrast.

Country	Before → after	Fixed	Broken
PE	13.4% → 93.7%	16,385	324
ES	48.0% → 91.5%	9,353	649
AR	80.3% → 59.7%	3,069	7,204
PH	94.7% → 54.9%	265	8,234

(The full table of 13 countries is in decisions/O6.)

Review

A design-docs review recomputed every figure and found eight problems; all are fixed and the corrections re-verified.

• A stale count. Two figures were copied from an agreement record made before Read the given names Marquis and Marquise as names outside a title phrase (#913) #923–Read the given name Queen as a name outside a title phrase (#913) #926 merged. The fresh figures on master are 249,526 of 357,517.
• Claims too broad for what was measured. "Every Spanish-speaking country": Argentina falls, and Uruguay's ratio is 7 to 1, not 8. "A Filipino middle name is the mother's surname": the measurement shows only how holders file it.
• A floor the sentence never named. The "17 countries under half" holds only for countries of at least 1,000 records.
• A confounded explanation. The Portugal contrast is presented as consistent with "who wrote the names down", not as a measure of it.

Argentina's after-figure is a rounding tie (11,930 of 20,000 = 59.65%), which the recipe prints as 59.7%.

Also done

• psc-individuals is re-recorded against master, as Add orcid-researchers, a local corpus of ORCID holders' own name boxes #927 asked. Scored with the scorer before and after Add orcid-researchers, a local corpus of ORCID holders' own name boxes #927 on the same tree, its 34,967 disagreeing records are identical entry for entry; only the fingerprint moved.

Checks

The doc tests pass, and both Sphinx builds (doctest and html) print nothing.

🤖 Generated with Claude Code