Should a maiden marker written with a colon read as one? (山田 花子 旧姓：佐藤 gives family 旧姓：佐藤, where 山田 花子 旧姓 佐藤 gives maiden 佐藤) #1359

Japanese writes a former surname as 旧姓：佐藤 or （旧姓：佐藤）, the marker and the name joined by a fullwidth colon. A colon is not a separator to rules/T1 or T2 (nor to any tokenize rule), so 旧姓：佐藤 is one word, the marker inside it is never looked up, and rules/M2 and M3 do not fire. rules/M2's Accepted clause and M3's Except record it; decisions/M2 and decisions/M3 each carry an Open: line for it pointing at #317, which is closed and was about Thai honorifics glued to the given name. #309, which first asked about 山田（旧姓：佐藤）, was closed in favour of #329, and #329 fixed the bracketed marker for the spaced form only. Both Open: lines, and rules/M2's "#317 tracks whether it should peel", should be given this issue's link.

Measured on master 670b349, 2026-11-12:

No row reports an ambiguity. Two things go wrong in the unbracketed Japanese rows beyond the missing maiden name: the marker word becomes the family name, and the name loses its script order. 山田 花子 旧姓：佐藤 reads given-first because the word holding the colon is not wholly Han, so rules/W4's family-first order for a wholly-Han name does not apply.

Expected: maiden 佐藤 in every Japanese row and maiden Jones in every Latin one. The colon is punctuation between a label and its value; no name holds one.

How often (2026-11-12): the corpora held cannot say. No local corpus is Japanese text of this kind: none of the 16 holds a fullwidth colon or 旧姓 in any name. 248 distinct names hold an ASCII colon at all (orcid-researchers 139, ctgov-officials 50, tse-donors 19, psc-individuals 18, fec-individuals 13, three others 9), and 2 of them, both in orcid-researchers, write a Latin maiden marker glued to a colon. The committed corpora hold 旧姓 in 7 names (5 in corpus_cjk.jsonl, 1 each from an issue and a rule) and the colon form once, rules/M2's own example. The form is standard in Japanese reference text (山田花子（旧姓：佐藤）); that is a claim about the convention, not a count.

What a colon is today. Nothing: rules/T1 splits at whitespace and ignorable characters, T2 at a script's own word divider, and neither lists : or ：. A colon stays inside its word, and a word holding one matches no vocabulary.

Options:

• (A) A colon after a listed maiden marker divides it from the name.
  • When: a word is a listed birth marker, a colon (: or ：), and at least one more character; or a listed marker with a colon ending it, before another word.
  • Then: the marker is read as the marker and the rest as the first word of its clause, as if a space stood there; M2 and M3 then read it as they read the spaced form. The colon is dropped with the marker.
  • Because: the colon is written between a label and what it labels, and the marker's vocabulary carries the licence: only a word that is a listed marker before its colon is cut.
  • Except: a marker followed by a comma stays no marker (M3). A word with a colon whose head is no listed marker is left whole.
  • Accepted: a name that really begins née: or 旧姓： reads as a clause; none is known.
  • This is the "head-peel" decisions/M2 names, the same vocabulary-licensed cut rules/W6 makes for a glued title, at tokenize or in extract.
• (B) A colon is a separator everywhere (rules/T1). Simpler, and it also reads Smith: John, but it changes every name holding a colon: the 248 above, which are mostly labels typed into a name field (Contact: …), times of day and typos, would each gain a word boundary. That needs its own measurement, and Should a lone dash be a separator, so John Smith - Jr. reads family 'Smith' rather than '-'? #820 (a lone dash as a separator) is the same kind of question.
• (C) Only the fullwidth colon, only after a Han marker. The narrowest script-scoped cut (the fullwidth colon is CJK punctuation, as the nakaguro is CJK-only, rules/T2). It leaves née:Jones as it is, which is rare by the count above.
• (D) Leave it as the recorded Accepted clause, and repoint the Open: lines here.

Open:

1. (A), (B), (C) or (D)?
2. Under (A) or (C), where does the cut happen: tokenize (as W6's front peel) or extract, where the bracketed form is read? The bracketed row goes through M3's "marker-led clause" test, which sees one word today.
3. Does the unbracketed form need the script order restored (山田 花子 旧姓：佐藤 → family 山田)? Under (A) it follows from the cut, since no word then holds the colon.
4. Other label punctuation after a marker: 旧姓・佐藤 is already divided by the nakaguro; geb.: Müller?

🤖 Generated with Claude Code