Skip to content

Read the Armenian Տեր as a title at the head and part of the surname inside the name (#1225) - #1320

Merged
derek73 merged 5 commits into
masterfrom
fix/issue-1225-ter
Nov 11, 2026
Merged

derek73 merged 5 commits into
masterfrom
fix/issue-1225-ter

Conversation

@derek73

@derek73 derek73 commented Nov 11, 2026 •

Copy link
Copy Markdown
Owner

Closes #1225.

What changes

Տեր is two things in Armenian names:

  • a cleric's title, used for a married priest and also for a bishop (Տեր Հուսիկ Լազարյան);
  • the spaced form of the Տեր- surname prefix (Զավեն Տեր Եղիյան).

It takes the three readings freiherr has (rules/P4's Except, a title-particle): a religious title, a rank particle, and a given reading. The given reading carries only the particle's may-be-given default and moves no parse; it follows freiherr's precedent.

Input 1.4.0 – 3.0.1 now
Տեր Հուսիկ Լազարյան given Տեր, middle Հուսիկ title Տեր, given Հուսիկ
Զավեն Տեր Եղիյան middle Տեր, family Եղիյան family Տեր Եղիյան
Տեր Եղիյան, Զավեն family Տեր Եղիյան unchanged
Լևոն Տեր-Պետրոսյան family Տեր-Պետրոսյան unchanged

What moves

  • Armenian Wikipedia's people (2026-11-10): all 10 names writing Տեր as a word of their own move, each to the reading it means: 5 clergy, 3 spaced Տեր- surnames, and 2 transcribed Dutch ter.
  • Corpora: 1 of 7 names moves, Լևոն Տեր Պետրոսյան in jrc-names, from middle Տեր to family Տեր Պետրոսյան. The committed corpora don't move.
  • Snapshot: gains only the two new P11 examples.
  • Reports: the given reading moves no field in a 2,220-parse grid. 96 of those parses gain a particle-or-given report, in exactly the places freiherr reports.

Name matching

Matching moves mostly because the parse does: the spaced surname is now read whole as the family name.

Pair master now
Զավեն Տեր Եղիյան / Զավեն Եղիյան middle Տեր held on one side family particles-aside
Լևոն Տեր-Պետրոսյան / Լևոն Տեր Պետրոսյան family part-of-other family joined
Զավեն Տեր Եղիյան / Զավեն Մարկոս Եղիյան inconsistent (the middles Տեր and Մարկոս conflict) consistent, as Juan de la Vega / Juan Marcos Vega already is
  • Keys. One change comes with no parse move: a hyphen half that is a listed particle is no key, so Լևոն Տեր-Պետրոսյան loses տեր;լ. On master, every hyphenated Տեր- surname with that initial shared that key.
  • Opt-out settings. Verdicts move as they do for any listed particle. Without particles-aside, the bare-surname pair becomes inconsistent; without one-side, it becomes consistent.
  • Tests. tests/v2/test_match.py pins the outcomes, the key set and the default verdict. Each check fails with the three readings removed.
  • match_diff.py --base against 86650d1, with the test file's two Տեր pairs in its draw: 8,211 of 654,444 rows move, every one holding a Տեր name.
    • All 21 verdict moves go from inconsistent to consistent: the Մարկոս pair under 19 of the 21 settings, and the Լևոն pair under -part-of-other and -one-side.
    • No consistent pair is left without a shared key.
    • The bare-surname move under -particles-aside isn't in the draw, so I measured it with Matching directly.

Accepted costs

These are recorded in decisions/salutation-titles. Each shape was misread before this change too.

  • A spaced Տեր surname written first with no comma (Տեր Եղիյան Զավեն) is the clergy shape word for word, so it reads the title.
  • A title before the spaced surname joins the title run (Պարոն Տեր Պետրոսյան gives title Պարոն Տեր), as Mr. Freiherr von Berg does.
  • Armenian Wikipedia writes the hyphenated Տեր- 235 times and the spaced form 8 times, 5 of those the clergy title.

Rules

rules/P11 (a lowercase particle opening a family-first name) gains a Precedence bullet against P4's title-particle. A title-particle opening the name reads as the title, its title reading taken first: freiherr von Berg Hans and տեր Բրաաք Մենո under FAMILY_FIRST both read the title. This was already true of freiherr; the rule now says so.

Review

The design-docs reviewer ran twice: once on the change and once on each fix commit. Every finding is fixed:

  • the report count and where the reports fall;
  • the matching section's claims (a third pass);
  • the title-run cost;
  • P11's missing Precedence bullet;
  • a 1.4.0 claim.

🤖 Generated with Claude Code

derek73 and others added 3 commits November 10, 2026 19:18
…inside the name (#1225)

Տեր is a married priest's title (Տեր Հուսիկ Լազարյան) and, spaced, the
Տեր- surname prefix (Զավեն Տեր Եղիյան). It takes the three readings
freiherr has: a religious title, a rank particle, and a given reading
carrying only the particle's may-be-given default. At the head of the
name the title is read; inside it the particle chains into the surname;
before a family comma the surname keeps it.

All 10 Armenian Wikipedia names writing the word move to the reading
they mean; in the corpora one name moves, Լևոն Տեր Պետրոսյան, middle
'Տեր' to family 'Տեր Պետրոսյան'. A Տեր- surname written first without a
comma is the clergy shape and reads the title, the accepted cost.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…le-particle precedence (#1225)

The design-docs review measured 96 parses gaining a particle-or-given
report from the given reading, not one, and found a title before the
spaced surname (Պարոն Տեր Պետրոսյան) joining the title run, by P4's
title-particle Except, as Mr. Freiherr von Berg does. Both are recorded
as accepted costs, with a case row. A lower-case title-particle opening
a family-first name reads the title where P11's writing would fold it;
P11 now states that as a Precedence bullet, freiherr and տեր its
examples.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… earlier reading (#1225)

The fix commit's second review: the 96 reports fall in three places,
the empty head the largest (54), which the entry did not name; the grid
is 740 texts, 2,220 parses; and Դոկտոր Տեր Եղիյան's earlier readings
were not stated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@derek73 derek73 added vocabulary Fixed by a word-list entry: per-word vocabulary and language packs titles Titles and honorifics (rules/H) non-latin-script Names written in a script other than Latin (rules/W) particles Particles and surname prefixes (rules/P) labels Nov 11, 2026
@derek73 derek73 self-assigned this Nov 11, 2026
derek73 and others added 2 commits November 10, 2026 19:41
Name matching reads the particle list, so the spaced Տեր- surname now
sets Տեր aside against the bare surname, the spaced and hyphenated
spellings are joined, and Լևոն Տեր-Պետրոսյան loses the key տեր;լ.
tests/v2/test_match.py holds all three, each failing with the readings
removed. match_diff.py --base on master moved 0 rows of a draw holding
no Տեր pair; with the test pair in the draw, 5,250 of 654,402, every
one holding that pair's names, no verdict going inconsistent.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… loosen as de la Vega's (#1225)

The bullet's review: "no parse moves" was wrong (the outcomes follow the
spaced surname being read whole; only the lost hyphen-half key needs no
parse move), its one-verdict sentence held only for match_diff's draw
(by default a spaced Տեր surname is now consistent with another middle
name, and the bare-surname pair turns inconsistent without
particles-aside, as Juan de la Vega reads), and the 0-row figure had no
recipe. A parametrized test pins the default verdict beside its de la
Vega twin, and match_diff is re-taken with it in the draw: 8,211 of
654,444 rows, every verdict move inconsistent to consistent.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@derek73
derek73 enabled auto-merge (squash) November 11, 2026 03:56
@derek73
derek73 merged commit 213f90f into master Nov 11, 2026
14 checks passed
@derek73
derek73 deleted the fix/issue-1225-ter branch November 11, 2026 04:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

non-latin-script Names written in a script other than Latin (rules/W) particles Particles and surname prefixes (rules/P) titles Titles and honorifics (rules/H) vocabulary Fixed by a word-list entry: per-word vocabulary and language packs

Projects

None yet

1 participant