Phase 3: language packs that prefer or restrict their language

Part of #747. Needs phase 2 and the language tags from the classification sub-issues.

The semantics (decisions/vocabulary-model): the default loads every language. A pack prefers its language: it settles clashes where its language has a vocabulary reading. A strict pack also drops other languages' name-level readings. Packs are sets of languages, so combining them is order-independent. Vocabulary in a non-Latin script is always on, because it selects itself by script. A pack can outrank a reading derived by a shape rule (Ma. Cristina, where Ma. is María).

Open: how today's locales packs move onto this; region and script subtags beyond zh-Hans/zh-Hant; whether third parties can ship packs.

Pack-specific issues are linked here as sub-issues.

A first measured case for "a pack prefers its language": Nepali (ne) against the Hindi and Marathi vocabulary.

The nepal-candidates local corpus (#866) writes each of 125,331 local-election candidates in Devanagari beside the Election Commission's own romanization, so it shows two languages' vocabularies disagreeing about one word, with the writer's intent recorded beside it.

श्री (#870). hi.toml and mr.toml list it as a salutation (Shri, Mr.; #269), and in a Hindi or Marathi name that is right. In Nepali names it is also the first element of a given name:

	names	today	the Commission's romanization
spaced `श्री` opening the name	215	title in all 215	a name word in 201 (`Shree` 193, `Shri` 6, `Sri` 2), `Mr` in 2
`श्री` glued into the given name	386	given	—

So श्री कृष्ण श्रेष्ठ reads title श्री while its own romanization Shree Krishna Shrestha reads given Shree.

Buddha, Siddha, Gyani, Guru (#869). Title words in the default vocabulary (buddha/siddha religious in en.toml, gyani in pa.toml, guru the renunciate title of #346), and the opening element of common Nepali compound given names: Buddha Bahadur Tamang reads title Buddha, given Bahadur. Of the names opening with one of them, 172 read it as a title.

Why this is a pack question and not a default change. Both readings are right for some language: the default loads every language, so it can only pick one, and either pick breaks the other language's names. Decision 6 (decisions/vocabulary-model) is the mechanism for exactly this: a ne pack that prefers its language settles the clash where Nepali has a vocabulary reading, reading श्री and Buddha as given-name elements, while the default and a hi pack keep the salutation. The Phase 2 spec (#752) decides how the default gates title words that are also names (its D12); the Nepali cases may still belong to the pack rather than to the default gate, and that is recorded there.

What a ne pack would carry, from these two issues: given-name readings for श्री (spaced, opening a name) and for buddha, siddha, gyani, guru opening a compound given name, with श्रीमती (92 names, romanized Shreemati in 62 and Mrs/Ms in 11) left a salutation. The corpus's two scripts give the pack its own test: a Devanagari name and its romanization should read the same roles.

Dotted and joined Nepali surnames for the ne pack (from #864, closed by #872)

#872 keeps exactly two dotted single letters (B.K., K.C., G.C.) as a name word at the end of a name, by shape, in the default. Some spellings are out of a shape rule's reach and belong to a ne pack that prefers a family reading:

Without periods (Ram Bahadur KC, BK, Bk). kc and bk are in no default set, so these already read as names, but nothing marks them as surnames. A pack can give bk, kc, gc, bc and dc a family reading.
dc / D.C., a live credential in the default: 99,172 NPPES providers write DC (Read a surname written as two dotted initials as a name, not a credential (#864) #872), against 21 nepal-candidates pairs ending in it. The default keeps the credential, and ne should flip it.
A pair behind an ambiguous credential (Jane Doe MA R.T.), which rules/S3 records as an accepted cost in the default. Under ne the trailing pair should be the family.
A strict ne pack would drop dc's credential reading outright.

Counts, nepal-candidates (Election 2079, 126,981 name pairs), pairs whose romanized name ends in the word, any spelling: bk 2,168, kc 497, gc 79, bc 70, dc 21. K.C. alone is written 341 times, K.C 126 and Kc 21. mc ends 2 pairs and gd none, so neither is on the list.

Measure the pack against nepal-candidates' script pairs, comparing roles rather than text: the Devanagari spelling says which word is the surname. That is the spelling-consistency measure the phase 2 spec (#752) uses.

Phase 3 input: words kept out of the default vocabulary (#753)

Swept 2026-10-27 on master cfe8bf92, read-only. Sources: every excluded entry in nameparser/vocab/*.toml; docs/design/decisions/ (Excluded, Declined and Open blocks, removal records, and every mention of a pack or an opt-in); PR #132; #18, #99 and the closed issues and PRs that list words; the open issues #658, #514, #804, #317, #345, #684, #725, #870 and #913, and #753's comments; the four Python locale packs. No bearer lookups were run: every count below is copied from its source.

The two figures

Words whose readings already clash between languages: 9, the same as the 2026-10-27 count.

8 clash by category, a different category in different files: abd (ar bound_given, en suffix), i (ca/pl conjunction, de/mul generation), st (cs generation, en/fr particle and title), ra (de title, en suffix), sr (en suffix, es/pt title), do (en credential, pt particle), mc (en particle and credential, ga/gd particle), vd (en credential, nl particle).
1 clashes by kind only, inside the particle category: de (a preposition in es/fr/it/pt, an article in nl). Prefer or strict changes nothing for de, since both kinds compile into PARTICLES.

Of the 110 words that appear in more than one file, these are the only clashes.

Readings with default = false: 3. They are cs st (generation, starší), de i (generation, the regnal ordinal written with its period) and es h (generation, hijo). All three are generation words that a period makes look like an initial (rules/T5).

What the inventory says

Most candidates are not about clashes. The 9 clashing words are nearly all a pack can settle under prefer. The (a) rows below are mostly readings that are missing from the default entirely, so the main job of phase 3 is to turn readings ON, not to settle clashes. That supports what the orchestrating session found.
Mechanisms: most (a) rows need default = false (DF, 22) or default = false plus strict (DF+ST, 27). DF+ST is the case where the reading is right in its own language and its collision is with another language's name: pan (Chinese surname Pan), bay (a Hungarian surname), nic (English Nic), puan (an Indonesian given name), fil (a Slavic surname), and the Dutch an/in/to/ten.
Prefer alone is small. Only 3 (a) rows (ms bt against the en credential Bt, cy ab against the en military title, cs bc against ne bc) plus the PO+PR and DF+PR mixes are settled by prefer as the RFC defines it.
Same-language clashes cannot be settled by any pack (the fils argument). Most of the 189 (c) rows are this kind:
- the Indic and Arabic title words that are also given names or surnames in their own language (कुमारी, ठाकुर, سلطان, سيد);
- ông, segundo, thiru, malam.
For several of these, the phase 2j route (a given reading plus the H9 title phrase, as sheikh, sultan and king took) is a better fit than a pack. I marked those "phase 2j route".
The (b) rows are 156:
- plurals (herren, heren);
- unattested words (bon, dí, heer, gravos);
- footnote marks (cva, gaee, np);
- removed credentials whose name sense dominates everywhere (rai, cha, mai);
- CLDR label translations and adjectives written before the name ("Junior" in more languages: Alexandre Dumas fils reads family fils #679);
- glued forms that need no particle (gil, kil, m', pour);
- the case-mask and attached exclusions, whose collision is in the word's own script.

Languages with enough candidates to justify a pack

ne: yes, and the evidence is the strongest of any language.
- Family readings for bk (2,168 candidate names), kc (497), gc (79), bc (70) and dc (21), with dc flipped from the en credential.
- A given reading for the spaced श्री (201 of 215 names).
- guru/गुरु, which Derek already assigned to a ne pack (2026-10-24).
Every one of these is a name reading. default = false is not allowed on the name kinds, so this pack needs a mechanism today's model does not have (Q1).
hi/bn (Ship hi and bn locale packs for Latin-transliterated Indic honorifics #345): yes, for the Latin transliterations: smt, shrimati, sushri, shriyut, shri, mst, adhyapak (DF+ST). About 30 more rows are (c) because the word is a name in the same language: kumari, babu, begum, acharya, sardar, pandit, sant, and the Tier 4 deferrals of decisions/indic-honorifics.
The Czech, Slovak, Slovenian and Croatian group: yes, but small. ml/st (generation; cs st is already DF while cs ml and the hr/sk/sl st are excluded instead, Q4), pan, roz, bc, and possibly mudr/phdr/mgr (Should a bare Ing after a comma be a credential? Viera Sevcikova, Ing reads given Ing, family Viera Sevcikova #848).
ga/gd: yes, small. nic (decisions name "an opt-in Irish pack" as its home), mhic/mic/uí, and the unaccented ni/ui. o, na, óg and mór are (c).
nl (and de): a small (a) set, all of the same shape. These are the particles that were confirmed regressions globally (an, in, ten, then, to, v; PR Add German/Dutch prefixes and German title/degree suffixes #191). Strict only removes the Korean and Vietnamese readings that break them (Park In Hwan, Nguyen To Nga) if those readings are recorded as name readings in the vocabulary (Q2).
pt-BR: yes, small. The single-letter agnomens F/N (Should José Silva Júnior read Júnior as a suffix, as José Silva Junior does, and where do Filho, Neto and Sobrinho belong? #843, which leaves them to phase 3 explicitly), fil and sb (DF+ST and DF+PR), and the unvetted abbreviations.
cy: a candidate list, never vetted. It comes from PR Tussenvoegsels / family name prefixes support #132: ap, ab, ferch spellings. Some may be safe in the default.
nb/sv: 36 (c) rows, all from PR Tussenvoegsels / family name prefixes support #132. These are the farm-name position words (nord, vestre, øvre…). Nobody ever judged them, and a Norwegian registry measurement would have to come first.
Every other language has 1 to 3 candidates. That is not enough for a pack on its own, but a pack file costs nothing once the mechanism exists:
- tr bay, fr père, it sig, lt jaun, mi a, br ar;
- ms bt/bte/puan;
- es lo;
- ber aït.

The locale packs today

ZH carries 556 surnames (Simplified, Traditional, ranked and compound forms). They are a segmentation list, outside the per-word model, like KOREAN_SURNAMES. Its keep-outs are recorded in decisions: married double surnames (李林), Japanese surnames (佐藤), and unattested compounds (东野, 成公, 宗政).
JA, RU and TR_AZ carry empty lexicons.
- RU's and TR_AZ's marker words live inside regexes in _pipeline/_post_rules.py: whole-word oğlu, qızı, kızı, uulu, оглы, кызы, улы…, and the endings -ovich/-ович…
- None of the Turkic markers is in the TOML, so moving them is new vocabulary, not a move (Q8).

Summary

411 candidate readings: 150 from the TOML excluded entries (102 word-level, 48 property-level) and 261 from decisions/, closed and open issues and PRs (one row per word-reading; a few spellings of one decision are rows of their own).

By verdict: (a) 66, (b) 156, (c) 189

toml: (a) 25, (b) 72, (c) 53
other: (a) 41, (b) 84, (c) 136

Pack candidates (a) by language, with mechanism:

language	(a)	(c)	mechanisms of the (a) rows
ne	8	0	PO 6, PO+PR 2
ga	6	4	DF+ST 3, DF 3
cy	6	0	DF 4, DF+ST 1, PR 1
hi	5	17	DF+ST 5
nl	5	1	DF+ST 5
cs	4	6	DF+ST 2, DF 1, PR 1
pt-BR	4	5	DF 2, DF+ST 1, DF+PR 1
ms	3	4	DF+ST 1, PR 1, DF 1
bn	2	6	DF+ST 2
ber	2	0	DF 2
hr	2	0	DF 2
pl	2	0	DF+ST 1, pack removes a property 1
sk	2	0	DF 2
sl	2	0	DF 2
ar	1	11	DF+ST 1
de	1	11	DF+ST 1
es	1	8	DF+ST 1
sv	1	8	DF-prop+ST 1
it	1	7	DF+ST 1
fr	1	4	DF 1
pt	1	4	pack removes a property 1
br	1	0	DF+ST 1
ca	1	0	pack removes a property 1
is	1	0	DF-prop+ST 1
lt	1	0	DF 1
mi	1	0	DF 1
tr	1	0	DF+ST 1
nb	0	27
en	0	16
bn/hi	0	9
?	0	7
bo	0	5
he	0	3
hu	0	3
th	0	3

(languages with no (a) and fewer than three (c) rows are left out of this table; every row is in the full table)

(a) rows by mechanism: DF+ST 27, DF 22, PO 6, PR 3, pack removes a property 3, DF-prop+ST 2, PO+PR 2, DF+PR 1

Questions for the Phase 3 design

Can a pack carry a name reading? Every ne candidate (bk, kc, gc, bc, dc, spaced श्री, guru's given reading) is a family or given reading. default = false is refused on the name kinds, and in the default a name reading is not inert: it gates a clashing reading or marks a fork. The design needs either a pack-only file that the default does not load, or default = false on name readings.
Strict sees only recorded name readings. DF+ST is the largest group (27 rows). In it, strict is supposed to remove the other language's name, for example Pan, In, To, Nic, Bay, Sig and Puan. It can do that only if that name is recorded as a name reading in some file. Several are not (for example in and to as Korean and Vietnamese given names). Does phase 3 need a sweep that records those name readings first, or should the pack carry the reading ungated?
Properties, not only readings. 48 exclusions refuse a property on a reading the default keeps: after_given, attached, cased, may_be_initial. The one-case connective fork (e pt, i ca/pl, y nl) is something a pack is meant to remove, and decisions/3-0-reevaluations records that a pack "cannot express 'remove e'". Should default = false (or its pack-side opposite) apply to a property, and can a pack remove one?
One shape, two encodings. cs st is default = false, while cs ml and the hr/sk/sl st and ml are excluded as "pack-shaped". Should phase 3 convert pack-shaped exclusions into default = false readings, so that an excluded reading again means "wrong everywhere"? The sweep would make that distinction mechanical: every (a) row in the table is an exclusion that would become a reading.
How a tag filters. cs/sk/sl/hr share ml/st, and ms/id share puan, putra and putri, where a strict ms pack would drop the Indonesian readings it needs. Should pt match pt-BR, which holds every agnomen candidate? (Open under Phase 3: language packs that prefer or restrict their language #753.)
Latin transliterations under strict. The hi/bn Latin forms clash both with names elsewhere (Indonesian Sri) and with names in the same language (Indian Shri/Sri given-name elements). A strict hi pack settles only the first. Is the remaining same-language cost accepted per word, as graf and herra were?
Not every candidate is a pack question. About a dozen (c) rows wait on a rule, not a word:
- Hungarian úr follows the name;
- Thai นาย needs a head peel;
- French M. is initial-shaped;
- the de Dr. rer. nat. modifiers.
Others want the default's H9 route instead: earl, don and baron (Should Baron, Bishop, Judge and the other title words borne as names read as names outside a title phrase? (Baron Davis → title 'Baron', no given name) #913), and the Arabic and Indic given-name titles. Phase 4's [rules] table, or phase 2j-style batches, may be the right home for them rather than a pack.
What moves from the Python packs. ZH's surname list stays a segmentation list, and the Turkic and East Slavic markers live in regexes. Do they become TOML readings (a new patronymic reading on oğlu, and so on), or stay rule data under phase 4?
Measuring a pack. The snapshot and compare.py parse under the default Lexicon, so neither can show what a pack changes (decisions/3-0-reevaluations). The agreement records score CLDR names under their language's pack since Measure the CLDR corpus with each locale pack too #673, and nepal-candidates holds every ne candidate. Does each pack need its own local-corpus reading, the way ne would need nepal-candidates?

Method and limits

Verdicts are mine, taken from the source's own reason. Each (c) row names what would decide it. The mechanism labels mean:
- DF: a default = false reading the pack turns on;
- ST: strict, dropping another language's name readings;
- PR: prefer, settling a clash between vocabulary readings;
- PO: a reading only the pack carries, which no current mechanism expresses for name kinds (Q1);
- DF-prop: a property turned on.
Rows group a few spellings. The 261 non-TOML rows include some spellings of one decision as rows of their own (the five Tibetan titles, the CLDR adjectives), so the counts are of word-readings, not decisions.
Language tags are best guesses for PR Tussenvoegsels / family name prefixes support #132's untagged words. Eleven are ? and were judged (b) for having no stated language or use.
Skipped: the 20 GitHub rows and 58 decisions rows that duplicate a TOML exclusion were merged into it. Words that shipped since they were proposed are not counted. PR Tussenvoegsels / family name prefixes support #132 has 112 words: 31 are particles now, 3 are excluded particles (ben, nic, o), 3 exist only as another kind (ab, e, i), and 75 are absent.
Should Baron, Bishop, Judge and the other title words borne as names read as names outside a title phrase? (Baron Davis → title 'Baron', no given name) #913's remaining title words are not enumerated in decisions/. Only baron, earl and don appear here.
Local corpora were not read.

Full table, part 1 of 2

Sorted by language tag, then verdict.

word	tag	kind	source	why kept out	verdict	mechanism
a	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
alam	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
at	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
ath	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
bert	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
dele	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
jẹnereṣọn	?	generation	decisions/generational-words	a CLDR label translation	b	—
kichik	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
kiçi	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
kiçik	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
lu	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
mala	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
nin	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
oz	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
war	?	particle	PR #132	never judged: #132 was closed for #182's last_base view; no language or use given in the PR	b	—
кенже	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
кесе	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
кіші	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
نسل	?	generation	decisions/generational-words	a CLDR label translation	b	—
जेनेरेसन	?	generation	decisions/generational-words	a CLDR label translation	b	—
ज्यु	?	generation	#679; decisions/generational-words	an abbreviation nobody writes after a name	b	—
द्वितीय	?	generation	#679; decisions/generational-words	means "second"	b	—
धा	?	generation	#679; decisions/generational-words	an abbreviation nobody writes after a name	b	—
ਜੂ	?	generation	#679; decisions/generational-words	an abbreviation nobody writes after a name	b	—
இளைய	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
ಕಿರಿ	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
බාල	?	generation	#679; decisions/generational-words	an adjective written before the name	b	—
ᏔᎵᏁ	?	generation	#679; decisions/generational-words	means "second"	b	—
φον	?	preposition	#670 (open)	"a rare edge case for the cost they would carry"	c	—
вон	?	preposition	#670 (open)	"a rare edge case for the cost they would carry"	c	—
фон	?	preposition	#670 (open)	"a rare edge case for the cost they would carry"	c	—
פון	?	preposition	#670 (open)	"a rare edge case for the cost they would carry"	c	—
فون	?	preposition	#670 (open)	"a rare edge case for the cost they would carry"	c	—
ფონ	?	preposition	#670 (open)	"a rare edge case for the cost they would carry"	c	—
ቮን	?	preposition	#670 (open)	"a rare edge case for the cost they would carry"	c	—
ልጅ	am	generation	#679	collides with the title Lij	b	—
ben	ar	patronymic	vocab/ar.toml:406	Ben is a common given name standing exactly where a forward-joining particle claim acts: "Ahmad Ben Husain" reads middle 'Ben', and admitting it would take that token … — clash is with the English given name Ben; a strict ar pack (Maghrebi data) drops it	a	DF+ST
mohamad	ar	bound_given	#133; decisions/vocabulary-exclusions	a standalone given name; binding it would eat the middle name	b	—
ال	ar	article	vocab/ar.toml:400	normally attached to the word it determines, and too ambiguous standing alone (#269) — attached article; never a standalone word	b	—
و	ar	conjunction/may_be_initial	decisions/vocabulary-exclusions	caseless letter: the fork is inert	b	—
أفندي	ar	honorific	decisions/vocabulary-exclusions	Ottoman post-nominals; surname collision	c	DF
أمير	ar	nobility	vocab/ar.toml:141	Amir is a common given name — same-language given name; phase 2j route (given reading + H9) rather than a pack	c	—
الأمير	ar	style/nobility	decisions/vocabulary-exclusions	article forms never vetted (الشيخ ships)	c	—
السلطان	ar	style/nobility	decisions/vocabulary-exclusions	article forms never vetted (الشيخ ships)	c	—
السيد	ar	style/nobility	decisions/vocabulary-exclusions	article forms never vetted (الشيخ ships)	c	—
باشا	ar	honorific	decisions/vocabulary-exclusions	Ottoman post-nominals; surname collision	c	DF
بك	ar	honorific	decisions/vocabulary-exclusions	Ottoman post-nominals; surname collision	c	DF
د	ar	academic	vocab/ar.toml:89	the abbreviation د.: the edge-period fold leaves bare د, a single letter that would swallow initials (the bare-κ trap) — single-letter shape trap; decided by whether Arabic-script data writes initials at all	c	DF
سلطان	ar	nobility	vocab/ar.toml:145	Sultan is a common given name — same-language given name; Latin sultan took the phase 2j route	c	—
سيد	ar	style	vocab/ar.toml:57	Sayyid is a common given name; the honorific is not shipped bare — same-language clash (Sayyid given); the Latin sayyid route of phase 2j (given reading + H9 phrase) may fit better than a pack	c	—
شيخ	ar	style	vocab/ar.toml:61	bare, it collides with the common female given name Shaikha; الشيخ, with the article, ships — same-language clash (Shaikha); Latin sheikh/shaikh took the phase 2j route (#912), which a given reading on شيخ could copy in the default	c	—
ক	as	generation	vocab/as.toml:4	the abbreviation ক. is the single-letter initial shape — single-letter initial shape inside its own script	b	—
мал	be	generation	vocab/be.toml:4	мал. is also "figure" — ambiguous with a common noun abbreviation; decided by Belarusian data	c	DF
ait	ber	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; Kabyle "sons of" (Aït Ahmed)	a	DF
aït	ber	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; Kabyle "sons of" (Aït Ahmed)	a	DF
млада	bg	generation	vocab/bg.toml:30	Mlada is a borne given name, and under a family-first order the given name is the trailing word — same-language given name, trailing under family-first order	b	—
adhyapak	bn	salutation	#345 (open)	as smt	a	DF+ST
mst	bn	salutation	vocab/bn.toml:4	the transliteration of মোসাঃ collides with real names where the native script cannot — Bangladeshi romanized records write Mst. (Mosammat) routinely; the collision is with names elsewhere	a	DF+ST
বাবু	bn	honorific/attached	vocab/bn.toml:160	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
মহারাজ	bn	courtesy/attached	vocab/bn.toml:176	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
মো	bn	salutation/after_given	vocab/bn.toml:36	abbreviates Mohammad, a given name, as মোঃ does (#853) — abbreviates Mohammad (#853)	b	—
মোঃ	bn	salutation/after_given	vocab/bn.toml:32	abbreviates Mohammad, a given name, and is read as one where it leads (#853) — abbreviates Mohammad (#853)	b	—
সাহেব	bn	honorific/attached	vocab/bn.toml:168	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
begum	bn	salutation	#345; #606	declined for the default; #345 open for a pack; a trailing name element	c	—
কুমারী	bn	salutation	vocab/bn.toml:8	Kumari is a given name and a family name in this script, so the leading claim fails C-i (decisions/vocabulary-collisions) — same-language given and family name; fails C-i inside Bengali	c	—
ঠাকুর	bn	style	vocab/bn.toml:72	a genuine Bengali honorific (lord, master) that is also Tagore, the surname — same-language surname (Tagore)	c	—
বেগম	bn	salutation	vocab/bn.toml:16	Begum is a trailing name element far more often than a leading title — same-language trailing name element	c	—
শেখ	bn	style	vocab/bn.toml:76	a clan name and a family name ('শেখ হাসিনা'); the Latin sheikh cluster ships for the Arabic addressing form, while the Indic spellings name the family — same-language clan and family name	c	—
শ্রী	bn	salutation/after_given	vocab/bn.toml:56	also an element of given names (Sri, Shri), so where it ends the given part it may be one — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
advocate	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
hafiz	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
marhum	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
shahid	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
syed	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
मां	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
माता	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
প্রকৌশলী	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
মা	bn/hi	title	decisions/indic-honorifics Tier 4	"skipped pending a corpus check, not declined"	c	—
rinpoche	bo	courtesy/attached	vocab/bo.toml:8	spaced-only for the reason the Indic trailing set is (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
geshe	bo	religious	#514 (open); decisions/indic-honorifics	per-word surname-retention questions; whether Latin forms belong in a pack	c	—
jetsun	bo	religious	#514 (open); decisions/indic-honorifics	per-word surname-retention questions; whether Latin forms belong in a pack	c	—
khenpo	bo	religious	#514 (open); decisions/indic-honorifics	per-word surname-retention questions; whether Latin forms belong in a pack	c	—
kyabje	bo	religious	#514 (open); decisions/indic-honorifics	per-word surname-retention questions; whether Latin forms belong in a pack	c	—
tulku	bo	religious	#514 (open); decisions/indic-honorifics	per-word surname-retention questions; whether Latin forms belong in a pack	c	—
ar	br	article	#658 (open)	proposed ambiguous like van; US surname, Turkish, Arabic ar-	a	DF+ST
i	ca	conjunction/may_be_initial	PR #132; decisions/3-0-reevaluations; decisions/P8	a ca or pl caller removes it from conjunctions_ambiguous by hand	a	pack removes a property
bc	cs	academic	#848 Q1	open whether Bc. follows Ing; bc is also a ne surname (#753)	a	PR
ml	cs	generation	vocab/cs.toml:21	pack-shaped: real and standard, but collides with ML initials in other languages written in Latin script — the cs st twin is already default = false; ml is excluded instead (see questions)	a	DF
pan	cs	salutation	vocab/cs.toml:4	Czech and Polish Mr., and the Chinese surname Pan, which leads; the accented paní ships — clash is with the Chinese surname Pan; a strict cs pack drops it	a	DF+ST
roz	cs	birth_marker	vocab/cs.toml:39	Roz, the English diminutive of Rosalind, stands exactly where a marker acts: with it listed, 'Rosalind Roz Smith' reads maiden 'Smith' and takes the given name Rosalin… — clash is with English Roz; within Czech the abbreviation still faces the marker-takes-everything cost	a	DF+ST
mgr	cs	academic	#848 Q1	open whether Mgr. follows Ing	c	—
mudr	cs	academic	#848 Q1	open whether they follow Ing (ambiguous credential)	c	—
paní	cs	salutation/after_given	vocab/cs.toml:8	spelled PANI, a surname (Census 2010), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
phdr	cs	academic	#848 Q1	open whether they follow Ing (ambiguous credential)	c	—
rodina	cs	salutation	vocab/cs.toml:13	Czech 'family', and a Russian surname, often written surname-first; familie ships — Russian surname clash is cross-language, but whether rodina leads a name as a title in Czech records is unmeasured	c	DF+ST
rozený	cs	birth_marker	decisions/vocabulary-exclusions	"awaits the same vetting, on its own merits"	c	—
ab	cy	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; ab ships as an en military title	a	PR
ap	cy	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; AP is also two initials	a	DF+ST
erch	cy	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; ferch "daughter of" spellings; no collision stated, so possibly default-safe	a	DF
fetch	cy	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; ferch "daughter of" spellings; no collision stated, so possibly default-safe	a	DF
verch	cy	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; ferch "daughter of" spellings; no collision stated, so possibly default-safe	a	DF
vetch	cy	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; ferch "daughter of" spellings; no collision stated, so possibly default-safe	a	DF
f	da	birth_marker	vocab/da.toml:36	the abbreviation of født collides with the initial F., so only the full participles født, fødd and född are safe — initial shape like es h, but a marker takes every word after it, so the cost of a false hit is larger than h's	c	DF
sonder	da	particle	PR #132	never judged: #132 was closed for #182's last_base view	c	DF+ST
v	de	preposition	#18 (gist); PR #191	bare v a confirmed regression; ships only as a numeral	a	DF+ST
herren	de	salutation	vocab/de.toml:30	the plural of Herr: a plural addresses several people (Herren Müller und Schmidt) and never stands before one person's name, so as a title it only took the first word … — plural; addresses several people (#766)	b	—
fürst	de	nobility (title-particle)	decisions/H6 Open #709	whether they read as freiherr before a family comma	c	—
graf	de	nobility (title-particle)	decisions/H6 Open #709	whether they read as freiherr before a family comma	c	—
herr	de	salutation/after_given	vocab/de.toml:17	a surname (Census 2010 HERR), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
hr	de	salutation/after_given	vocab/de.toml:26	ending a given part in US data, more likely two initials run together than the abbreviation of Herr: one of three such fec-individuals names has a woman's given name (… — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
mag	de	academic	#18	mag ships as an en office; Mag. reads by shape	c	PR
med	de	modifier	#18; #41	never addressed; dotted forms already read by H2's shape; Nat and Phil are given names	c	—
nat	de	modifier	#18; #41	never addressed; dotted forms already read by H2's shape; Nat and Phil are given names	c	—
phil	de	modifier	#18; #41	never addressed; dotted forms already read by H2's shape; Nat and Phil are given names	c	—
prinz	de	nobility (title-particle)	decisions/H6 Open #709	whether they read as freiherr before a family comma	c	—
rer	de	modifier	#18; #41	never addressed; dotted forms already read by H2's shape; Nat and Phil are given names	c	—
univ	de	modifier	#18; #41	never addressed; dotted forms already read by H2's shape; Nat and Phil are given names	c	—
gravos	el	nobility	#606; decisions/salutation-titles	not shown to be the Greek for Count	b	—
κ	el	salutation	vocab/el.toml:4	κ. (Mr.) has the initial-and-surname shape: 'Κ. Παπαδόπουλος' would read title 'Κ.' with an empty given, the edge-period fold making the entry match the initial; Latin… — single-letter initial shape inside Greek; es h / de i are the default = false precedent	c	DF
κόμης	el	nobility	#606; decisions/salutation-titles	a Greek surname (~420 bearers); "a decision still to make"	c	—
acca	en	credential	#35	surnames or company text (derek73, 2015)	b	—
administration	en	title	#606	out of scope: estate and organisation detection	b	—
agente	en	credential	#35	surnames or company text (derek73, 2015)	b	—
born	en	birth_marker	decisions/vocabulary-exclusions	never shipped; Max Born	b	—
cha	en	credential	vocab/en.toml:2948	Cha, the Korean 차, dominates CHA (Certified Hotel Administrator): removed 2026-09-07 by the relative-frequency criterion (decisions/suffix-acronym-collisions) — removed: the Korean surname dominates even in English data (relative-frequency criterion)	b	—
commission	en	title	#606	out of scope: estate and organisation detection	b	—
cva	en	credential	decisions/suffix-acronym-collisions	Wikipedia footnote marks	b	—
ded	en	credential/cased	vocab/en.toml:3492	a mask applies in every role, and Ded is an Albanian given name: 'ded gjo luli' would render 'DEd Gjo Luli' (measured 2026-09-24) — a mask applies in every role; not language-bound	b	—
department	en	title	#606	out of scope: estate and organisation detection	b	—
do	en	academic	PR #32	"someone named Do" (derek73, 2015)	b	—
edd	en	credential/cased	vocab/en.toml:3620	a mask applies in every role, and Edd is borne as a given name: 'edd smith' would render 'EdD Smith' and 'EDD KIMBER' 'EdD Kimber' (measured 2026-09-24) — a mask applies in every role; not language-bound	b	—
esq	en	credential	decisions/vocabulary-exclusions	acronym readings removed (esq stays a word)	b	—
estate	en	title	#606	out of scope: estate and organisation detection	b	—
gaee	en	credential	decisions/suffix-acronym-collisions	Wikipedia footnote marks	b	—
gc	en	credential	vocab/en.toml:3857	G.C. (Gharti Chhetri) is a Nepali surname written as two dotted initials, ending 80 of 126,981 name pairs in Nepal's 2079 BS local election candidates, where GC as a c… — removed; the pack interest runs the other way: ne wants gc as a family name (#864)	b	—
globally	en	credential	#35	surnames or company text (derek73, 2015)	b	—
hiring	en	credential	#35	surnames or company text (derek73, 2015)	b	—
lac	en	credential/cased	vocab/en.toml:4040	a mask applies in every role, and Lac is a Vietnamese given name and a French surname: 'tran lac' would render 'Tran LAc' (measured 2026-09-25, #540) — a mask applies in every role; not language-bound	b	—
leed ap	en	credential	decisions/vocabulary-exclusions; comma-suffix-arc	multi-word entries removed in 2.0	b	—
mai	en	credential	vocab/en.toml:4160	Mai, a Vietnamese given and family name, a Chinese surname (麥) and a Japanese given name, dominates MAI (Member, Appraisal Institute): removed 2026-10-16 by the relati… — removed: relative-frequency criterion	b	—
md	en	academic	vocab/en.toml:4241	a leading Md abbreviates Muhammad, a given name: 1,246 of 1,300 UK PSC filers with Md first in the forename box typed it there alone (2026-10-23), and a title reading … — leading Md is Muhammad (#853)	b	—
meng	en	credential/cased	vocab/en.toml:4258	a mask applies in every role, and Meng is borne as a given name and a surname: 'MENG LI' would render 'MEng Li' and 'meng wang' 'MEng Wang' (measured 2026-09-24) — a mask applies in every role; not language-bound	b	—
nicet i	en	credential	decisions/vocabulary-exclusions; comma-suffix-arc	multi-word entries removed in 2.0	b	—
np	en	credential	decisions/suffix-acronym-collisions	Wikipedia footnote marks	b	—
ph	en	credential	decisions/vocabulary-exclusions	acronym readings removed (esq stays a word)	b	—
pi	en	credential	vocab/en.toml:4582	PI is Principal Investigator, a role and not a credential (102 ClinicalTrials.gov study officials write it after a comma, data 2026-10-21; 8 NPPES providers), and Pi i… — a role, not a credential (#797)	b	—
psm i	en	credential	decisions/vocabulary-exclusions; comma-suffix-arc	multi-word entries removed in 2.0	b	—
rai	en	credential	vocab/en.toml:4681	Rai, a common surname across Hindi- and Bengali-speaking regions, dominates RAI (RETA Authorized Instructor): removed 2026-09-07 by the relative-frequency criterion (d… — removed: relative-frequency criterion	b	—
sa	en	military	vocab/en.toml:784	Seaman Apprentice, with no bearer in any corpus: every name read with SA as its title in the committed and local corpora is a person whose given name is Sa (Sa Liu, Mo… — no bearer (#857)	b	—
sa	en	credential	vocab/en.toml:784	no person in the corpora writes it: the corpora's SA and S.A. suffixes are companies (Galderma S.A.), the South African qualifier in FCP(SA), and the initials S.A. typ… — no writer (#857)	b	—
senior	en	generation	vocab/en.toml:2198	Senior is a surname (3,864 US Census bearers) — English surname; same-language	b	—
wm	en	title	vocab/en.toml:4	William, abbreviated, and no title of any kind: #68 asked that 'Wm. Scott Magargee III' not read given name 'Wm.', a request judged mistaken, since Wm is the given nam… — a given name abbreviation (#789), never a title	b	—
ass	en	credential/modifier	#804	out of scope: "a separate decision"	c	—
bachelor	en	credential/modifier	#804	out of scope: "a separate decision"	c	—
baron	en	nobility	decisions/H1 Open #913	Baron Davis reads title; not yet gated	c	— (H9 default)
beng	en	credential	#18 (gist)	never addressed (meng ships ambiguous)	c	—
doctorate	en	credential/modifier	#804	out of scope: "a separate decision"	c	—
don	en	nobility	#913 Q4 (open); decisions/salutation-titles (#847)	proposed phrase-only like dean; borne names	c	— (H9 default)
duke	en	nobility	vocab/en.toml:367	a common given name (Duke Kahanamoku), and a surname besides — English given name; the phase 2j given-reading + H9 route, not a pack	c	—
earl	en	nobility	#913 Q4 (open); decisions/salutation-titles (#847)	proposed phrase-only like dean; borne names	c	— (H9 default)
fitz	en	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; usually glued (FitzGerald); Fitz is a given name	c	—
goodman	en	salutation/after_given	vocab/en.toml:9	a common surname (Benny Goodman), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
master	en	salutation/after_given	vocab/en.toml:26	a surname and a given name, borne as a middle name either way — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
medicine	en	credential/modifier	#804	out of scope: "a separate decision"	c	—
mister	en	salutation/after_given	vocab/en.toml:44	a given name and a surname, borne as a middle name either way — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
mz	en	salutation	#114 comment	never addressed	c	—
specialist	en	credential/modifier	#804	out of scope: "a separate decision"	c	—
venerable	en	religious (given-name title)	decisions/indic-honorifics	traditions split on surname retention; "deliberately NOT an Excluded entry"	c	—
lo	es	article	decisions/vocabulary-collisions (2026-08-17)	ambiguous: Lo leads romanized Chinese names (#360); never added	a	DF+ST
el	es	article	PR #132; #133; decisions/vocabulary-collisions (2026-08-17)	#360 left el "needing judgement"; lo judged ambiguous but never added	c	DF
profesor	es	academic	#804	open (123 ctgov officials)	c	—
profesora	es	academic	#606; #804	#804 open	c	—
segundo	es	generation	#843	never decided; a Spanish given name and surname	c	—
señor	es	salutation/after_given	vocab/es.toml:4	spelled SENOR, a surname (Census 2010), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
señora	es	salutation/after_given	vocab/es.toml:13	a given name (US SSA, as Senora), borne as a middle name — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
sr.ª	es	salutation	#840; #606	never addressed	c	—
y	es	conjunction/may_be_initial	vocab/es.toml:100	the commonest Hispanic double-surname link and this library's oldest fixture, 'Velasquez y Garcia': marking it would read the link of every one-case Hispanic record as… — "a caller who wants the Dutch reading adds it": an nl pack could carry it, but y is an es reading	c	—
härra	et	salutation/after_given	vocab/et.toml:4	spelled HARRA, a surname (Census 2010), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
pour	fa	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; written glued (-pour), and Pour is borne as a surname	b	—
پدر	fa	generation	vocab/fa.toml:28	"father" without the article that marks ar الابن and he הבן as never a name — bare noun "father", same-language	b	—
پسر	fa	generation	vocab/fa.toml:34	"son" without the article that marks ar الابن and he הבן as never a name — bare noun "son", same-language	b	—
herra	fi	salutation/after_given	vocab/fi.toml:4	a surname (Census 2010 HERRA), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
ft	fil	particle	#725	out of scope (under 100 Census bearers); mt ships as a credential	b	—
mt	fil	particle	#725	out of scope (under 100 Census bearers); mt ships as a credential	b	—
sta	fil	particle	#725 (open); decisions/H6	"weighed and not decided"	c	PR
sto	fil	particle	#725 (open); decisions/H6	"weighed and not decided"	c	PR
père	fr	generation	vocab/fr.toml:104	pack-shaped: collides outside French — "pack-shaped" in the source; its twin fils shipped gated in phase 2h	a	DF
m	fr	salutation	#658 (open)	M. is initial-shaped: "a rule, not a word list"	c	DF
pr	fr	academic	#804	open; no name readings found in US data	c	—
professeur	fr	academic	#804	open; no name readings found in US data	c	—
professeure	fr	academic	#804	open; no name readings found in US data	c	—
frou	fy	salutation	PR #99	never addressed	c	—
mhic	ga	clan	PR #132	never judged: #132 was closed for #182's last_base view; Irish clan forms beside the shipped mac/ní/ó; uí may be default-safe	a	DF
mic	ga	clan	PR #132	never judged: #132 was closed for #182's last_base view; Irish clan forms beside the shipped mac/ní/ó; uí may be default-safe	a	DF
ni	ga	clan	PR #132	never judged: #132 was closed for #182's last_base view; unaccented: Ni is a Chinese surname, ui an initial pair	a	DF+ST
nic	ga	clan	vocab/ga.toml:24	the Irish "daughter of" beside 'ní', but also an abbreviation of Nicolas and Nicole and a given name of its own (Nic Cage): a leading Nic would be marked ambiguous at … — source names "an opt-in Irish pack" as its home; the clash is with English Nic	a	DF+ST
ui	ga	clan	PR #132	never judged: #132 was closed for #182's last_base view; unaccented: Ni is a Chinese surname, ui an initial pair	a	DF+ST
uí	ga	clan	PR #132	never judged: #132 was closed for #182's last_base view; Irish clan forms beside the shipped mac/ní/ó; uí may be default-safe	a	DF
gil	ga	clan	PR #132	never judged: #132 was closed for #182's last_base view; written glued (Gilchrist, Kilpatrick); Gil is a Spanish given name	b	—
kil	ga	clan	PR #132	never judged: #132 was closed for #182's last_base view; written glued (Gilchrist, Kilpatrick); Gil is a Spanish given name	b	—
sinsearach	ga	generation	vocab/ga.toml:4	a rank word, not a generation; Irish writes Óg and Mór — rank word, not a generation	b	—
sóisearach	ga	generation	vocab/ga.toml:10	a rank word, not a generation; Irish writes Óg and Mór — rank word, not a generation	b	—
mór	ga	generation	#661; decisions/generational-words	also inside given names (Seán Óg); "a separate decision and not taken here"	c	DF
na	ga	article	PR #132	never judged: #132 was closed for #182's last_base view	c	DF
o	ga	clan	vocab/ga.toml:40	bare O is a common periodless initial, so a particle reading would turn the middle initial of 'Liam O Murchú', and of every 'John O Smith', into a surname; the anglici… — bare O is a periodless initial in Irish names too; decided by how often Irish data writes unaccented, unglued O	c	DF

Full table, part 2 of 2

Sorted by language tag, then verdict.

word	tag	kind	source	why kept out	verdict	mechanism
óg	ga	generation	#661; decisions/generational-words	also inside given names (Seán Óg); "a separate decision and not taken here"	c	DF
m'	gd	clan	PR #132	never judged: #132 was closed for #182's last_base view; written glued (M'Donald, McKay)	b	—
mck	gd	clan	PR #132	never judged: #132 was closed for #182's last_base view; written glued (M'Donald, McKay)	b	—
shehu	ha	style	vocab/ha.toml:10	a common Hausa given name: the leading title claim took the given name of Shehu Shagari, who read title Shehu in every release from 1.4.0 through 2.4.1; CLDR's Hausa r… — removed (#660): same-language given name	b	—
malam	ha	salutation	vocab/ha.toml:4	Hausa teacher or Mr.: a decline on the counts, for the lower-bound given-name use, though CLDR's Hausa data files Malam as the title; whoever re-argues it counts Hausa… — same-language given name; the H9 route, re-argued on Hausa counts	c	—
imam	he	religious (given-name title)	decisions/indic-honorifics	kept out of the renunciate class (title stays)	b	—
rabbi	he	religious (given-name title)	decisions/indic-honorifics	kept out of the renunciate class (title stays)	b	—
בר	he	patronymic	vocab/he.toml:122	Bar is a common modern Israeli given name, and the surname spelling is hyphenated anyway (Bar-Lev) — same-language given name; the surname form is hyphenated	b	—
סניור	he	generation	vocab/he.toml:76	borne as a surname in Israel (אלי סניור, בוריס סניור) — same-language surname	b	—
bath	he	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; Hebrew "daughter of"/"house of"; bat ships	c	DF
bet	he	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; Hebrew "daughter of"/"house of"; bat ships	c	DF
רב	he	religious	decisions/vocabulary-exclusions	bare רב is the ordinary word "many"	c	—
shri	hi	salutation	vocab/hi.toml:4	the transliteration of श्री collides with real given names where the native script cannot — the native श्री ships; the Latin clash is with given names (Sri/Shri, Indonesian Sri) that strict hi drops only partly	a	DF+ST
shrimati	hi	salutation	#345 (open); decisions/vocabulary-exclusions	Latin transliterations collide (#269); "belong, if anywhere, to the opt-in packs of #345"	a	DF+ST
shriyut	hi	salutation	#345 (open); decisions/vocabulary-exclusions	Latin transliterations collide (#269); "belong, if anywhere, to the opt-in packs of #345"	a	DF+ST
smt	hi	salutation	#345 (open); decisions/vocabulary-exclusions	Latin transliterations collide (#269); "belong, if anywhere, to the opt-in packs of #345"	a	DF+ST
sushri	hi	salutation	#345 (open); decisions/vocabulary-exclusions	Latin transliterations collide (#269); "belong, if anywhere, to the opt-in packs of #345"	a	DF+ST
जी	hi	honorific/attached	vocab/hi.toml:181	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
महाराज	hi	courtesy/attached	vocab/hi.toml:205	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
लामा	hi	religious	#514; decisions/indic-honorifics	too rare (#346)	b	—
साहब	hi	honorific/attached	vocab/hi.toml:189	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
साहिब	hi	honorific/attached	vocab/hi.toml:197	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
acharya	hi	religious	#345 (open)	a Brahmin surname, same language	c	—
babu	hi	salutation	#345 (open)	also given names in the same language	c	—
kumari	hi	salutation	#345 (open)	also given names in the same language	c	—
pandit	hi	academic	vocab/hi.toml:65	a common surname (Vijaya Lakshmi Pandit); the native पंडित ships — same-language surname; native पंडित ships	c	—
pt	hi	academic	#345	Pandit abbreviated; pt ships as an en credential	c	PR
sant	hi	religious	vocab/hi.toml:109	a Punjabi given name (Sant Singh); the native संत ships — same-language (Punjabi) given name; native संत ships	c	—
sardar	hi	salutation	#345 (open)	a Sikh given name and surname, same language	c	—
sri	hi	salutation	vocab/hi.toml:8	a given name (Sri Mulyani); the native श्री cannot collide — Sri is also an Indian given-name element, so even strict hi keeps a same-language clash	c	DF+ST
आचार्य	hi	religious	vocab/hi.toml:117	Acharya is a Brahmin surname — same-language surname	c	—
कुमारी	hi	salutation	vocab/hi.toml:12	Kumari is a given name and a family name in this script, so the leading claim fails C-i (decisions/vocabulary-collisions) — same-language given and family name	c	—
ठाकुर	hi	style	vocab/hi.toml:49	Tagore, the surname, as with the Bengali ঠাকুর — same-language surname	c	—
बाबू	hi	salutation/after_given	vocab/hi.toml:20	romanized Babu, an element of compound given names in India (Ram Babu) and a surname — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
बेगम	hi	salutation	vocab/hi.toml:25	Begum is a trailing name element far more often than a leading title — same-language trailing name element	c	—
राजा	hi	nobility	vocab/hi.toml:153	Raja is a common given name — same-language given name; phase 2j route	c	—
रानी	hi	nobility	vocab/hi.toml:157	Rani is a common given name — same-language given name; phase 2j route	c	—
शेख	hi	style	vocab/hi.toml:57	a clan name and a family name; the Latin sheikh cluster ships for the Arabic addressing form, while the Indic spellings name the family — same-language clan and family name	c	—
श्री	hi	salutation/after_given	vocab/hi.toml:29	also an element of given names (Sri, Shri), so where it ends the given part it may be one — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
ml	hr	generation	vocab/hr.toml:16	pack-shaped: real and standard, but collides with ML initials in other languages written in Latin script — as cs ml	a	DF
st	hr	generation	vocab/hr.toml:30	pack-shaped: real and standard, but St. is Saint in other languages written in Latin script — cs st is already default = false; these three are excluded instead	a	DF
id	hu	generation	vocab/hu.toml:8	written before the name, and a bare id would make any trailing ID a suffix — written before the name; bare id would make any trailing ID a suffix	b	—
ifj	hu	generation	vocab/hu.toml:14	written BEFORE the name (ifj. Kovács János), which H2's shape rule already reads as a title — written before the name, where H2 already reads ifj. as a title	b	—
asszony	hu	salutation	#606	follows the name (Kovács úr); bare trailing titles never claim (rules/H5)	c	rule (phase 4)
gróf	hu	nobility	vocab/hu.toml:4	Hungarian Count: possibly a surname in a surname-first culture; unverified, and under uncertainty the answer is do not ship — unverified as a Hungarian surname; a measurement decides it	c	DF
úr	hu	salutation	#606	follows the name (Kovács úr); bare trailing titles never claim (rules/H5)	c	rule (phase 4)
ավագ	hy	generation	vocab/hy.toml:16	an Armenian given name, in lists that write the surname first — same-language given name	b	—
paron	hy	salutation	vocab/hy.toml:4	Armenian Mr. in Latin script: unverified, so not shipped — unverified; if verified it may need no pack	c	—
tikin	hy	salutation	vocab/hy.toml:8	Armenian Mrs. in Latin script: unverified, so not shipped — unverified; if verified it may need no pack	c	—
setia	id	particle	PR #132	never judged: #132 was closed for #182's last_base view; given-name elements	b	—
setya	id	particle	PR #132	never judged: #132 was closed for #182's last_base view; given-name elements	b	—
frú	is	salutation/after_given	vocab/is.toml:4	spelled FRU, a Cameroonian surname, as fru is (sv.toml), and middle names are drawn from surnames — surname evidence is Cameroonian, another language	a	DF-prop+ST
sig	it	salutation	vocab/it.toml:4	Sig is borne as a given name (US SSA 50+ male, Wikidata 11+ given, Norway and the United States; the FEC contributors write 'SMITH, SIG' 16 times), measured 2026-10-22… — the clash is Norwegian/US given name Sig; Sig. already reads by shape	a	DF+ST
avv	it	academic	#840	bare forms "if vetting supports them"; dotted forms read by shape	c	DF
dott	it	academic	#840	bare forms "if vetting supports them"; dotted forms read by shape	c	DF
dott.ˢˢᵃ	it	salutation	#840	never addressed (ordinal-indicator spellings)	c	—
egr	it	academic	#840	bare forms "if vetting supports them"; dotted forms read by shape	c	DF
sig.ª	it	salutation	#840	never addressed (ordinal-indicator spellings)	c	—
signor	it	salutation/after_given	vocab/it.toml:21	a surname (Census 2010 SIGNOR), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
signore	it	salutation/after_given	vocab/it.toml:16	a surname (Census 2010 SIGNORE), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
シニア	ja	generation	vocab/ja.toml:4	left out (Derek, 2026-10-14) beside its Hebrew twin; its only count is Wikidata's Senior surname label, 96 bearers, which measures the English surname — left out by Derek 2026-10-14	b	—
フォン	ja	preposition	#671	also Fong	b	—
佐藤	ja	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	Japanese surnames belong to locales.JA's segmenter	b	—
博士	ja	credential/attached	vocab/ja.toml:32	glued 田中博士 IS Tanaka Hiroshi, an attested given name — the clash is inside the word's own script and language	b	—
殿	ja	honorific/attached	vocab/ja.toml:76	some ninety Japanese surnames end in it, 鵜殿 (Udono) and 真殿 (Madono) with four-figure populations, so a peel would cut a family name in two — the clash is inside the word's own script and language	b	—
氏	ja	honorific/attached	vocab/ja.toml:84	王氏 is a historical name form ("the Wang woman") — the clash is inside the word's own script and language	b	—
部長	ja	professional_honorific/attached	vocab/ja.toml:133	長 alone is a borne given name, so 阿部長 can be 阿部 + 長 (#707). The measurement found no such bearer, as it found none for the words that shipped; the refusal rests on the… — the clash is inside the word's own script and language	b	—
鈴木	ja	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	Japanese surnames belong to locales.JA's segmenter	b	—
馬丁	ja	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	Japanese surnames belong to locales.JA's segmenter	b	—
джуниор	kk	generation	vocab/kk.toml:4	a given name or a club name there — given or club name	b	—
ជូនៀ	km	generation	vocab/km.toml:4	a given name; ជូនៀរ ships — given name; ជូនៀរ ships	b	—
교수	ko	honorific	vocab/ko.toml:23	reads as a common noun as readily as an address term; 교수님 ships — common noun; 교수님 ships	b	—
군	ko	honorific/attached	vocab/ko.toml:29	김지군 is a given name ending in it — the clash is inside the word's own script and language	b	—
선생	ko	honorific	vocab/ko.toml:43	reads as a common noun as readily as an address term; 선생님 ships — common noun; 선생님 ships	b	—
양	ko	honorific/attached	vocab/ko.toml:56	김지양 is a given name ending in it, and 양 is a top-tier surname besides — the clash is inside the word's own script and language	b	—
폰	ko	preposition	#671	as フォン	b	—
जेआर	kok	generation	vocab/kok.toml:4	rare, and also JR — rare	b	—
tre	kw	particle	PR #132	never judged: #132 was closed for #182's last_base view; a place-name prefix, glued	b	—
jaun	lt	generation	vocab/lt.toml:16	the abbreviation jaun., refused with the full words shipping — abbreviation of shipped jaunesnysis; source states no collision, only caution	a	DF
a	mi	conjunction	decisions/vocabulary-exclusions (#478)	the Māori link in Te Awanui-a-Rangi; a spaced a is an initial or a name word elsewhere	a	DF
јуниор	mk	generation	vocab/mk.toml:12	a given name or a club name there; помладиот ships — given or club name	b	—
жр	mn	generation	vocab/mn.toml:8	the initials ЖР — two-letter initials shape; es h precedent	c	DF
क	mr	generation	vocab/mr.toml:64	the abbreviation क. is the single-letter initial shape — single-letter initial shape	b	—
जी	mr	honorific/attached	vocab/mr.toml:78	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
साहेब	mr	honorific/attached	vocab/mr.toml:86	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
bt	ms	patronymic	vocab/ms.toml:43	the abbreviation of binti, not reviewed; and already a suffix acronym (Baronet) with a Bt case mask, so a particle reading would open a particle-suffix overlap (the cl… — clashes with the en credential Bt (Baronet); prefer ms settles it, but "not reviewed"	a	PR
bte	ms	patronymic	vocab/ms.toml:49	the abbreviation of binte, not reviewed — "not reviewed"; no stated collision	a	DF
puan	ms	salutation	vocab/ms.toml:12	Malay Mrs., and an Indonesian given name (Puan Maharani) — clash is the Indonesian given name; a strict ms pack drops id readings (but ms and id names overlap, see questions)	a	DF+ST
cik	ms	salutation	vocab/ms.toml:4	Malay Miss: unverified, so not shipped — unverified	c	—
putera	ms	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; Putra and Putri are common Malay/Indonesian given names	c	—
putra	ms	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; Putra and Putri are common Malay/Indonesian given names	c	—
putri	ms	patronymic	PR #132	never judged: #132 was closed for #182's last_base view; Putra and Putri are common Malay/Indonesian given names	c	—
ing	mul	academic	vocab/mul.toml:50	Ing is borne as a given name (nam_dict band 5 in Sweden, Wikidata 10+ given) and as a surname (US Census 1,814), measured 2026-10-22 with tools/name_usage.py, and a ti… — borne as a name; Ing. reads by shape; credential shipped ambiguous (#848)	b	—
aust	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
austre	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
lille	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
mellom	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
myljom	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
ned	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
neder	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
nedre	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
nord	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
nordre	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
ny	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
opp	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
ost	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
oster	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
over	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
ovre	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
ovste	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
sondre	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
sor	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
sore	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
stor	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
syd	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
syndre	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
vesle	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
vest	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
vestre	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
vetle	nb	particle	PR #132	never judged: #132 was closed for #182's last_base view; farm-name position words (north, upper, little) written before a farm surname	c	DF+ST
bc	ne	family	#753 comment 2026-10-25	70 candidate names; clashes with cs Bc. if that ships	a	PO+PR
bk	ne	family	#753 comment 2026-10-25	ne pack family reading (2,168 and 497 candidate names)	a	PO
dc	ne	family	#753 comment 2026-10-25	default keeps the credential (NPPES 99,172); ne flips it, strict ne drops the credential	a	PO+PR
gc	ne	family	#753; #864	79 candidate names; the en credential was removed	a	PO
guru	ne	given (title stays)	decisions/vocabulary-model 2026-10-24; #753; #869	"guru goes to a ne pack in phase 3"; the Devanagari twin moves with it	a	PO
kc	ne	family	#753 comment 2026-10-25	ne pack family reading (2,168 and 497 candidate names)	a	PO
गुरु	ne	given (title stays)	decisions/vocabulary-model 2026-10-24; #753; #869	"guru goes to a ne pack in phase 3"; the Devanagari twin moves with it	a	PO
श्री	ne	given	#870 (open); #753 comment 2026-10-24	a given-name element in 201 of 215 Nepali names; options include a ne/hi/mr pack	a	PO
जेआर	ne	generation	vocab/ne.toml:8	rare, and also JR — rare	b	—
an	nl	preposition	#18 (gist); PR #191	confirmed regressions as global particles (Park In Hwan, Nguyen To Nga)	a	DF+ST
in	nl	preposition	#18 (gist); PR #191	confirmed regressions as global particles (Park In Hwan, Nguyen To Nga)	a	DF+ST
ten	nl	preposition	#18 (gist); PR #191	confirmed regressions as global particles (Park In Hwan, Nguyen To Nga)	a	DF+ST
then	nl	preposition	#18 (gist); PR #191	confirmed regressions as global particles (Park In Hwan, Nguyen To Nga)	a	DF+ST
to	nl	preposition	#18 (gist); PR #191	confirmed regressions as global particles (Park In Hwan, Nguyen To Nga)	a	DF+ST
heer	nl	particle	vocab/nl.toml:49	a rank particle ('lord', the nobiliary 'heer van') added from #18's test "Johan de heer Van Kampen" (family 'de heer Van Kampen'), which appears to be two real names r… — no attestation (#766)	b	—
heren	nl	salutation	vocab/nl.toml:8	the plural of heer ('de heren'): a plural addresses several people and never stands before one person's name, so as a title it only took the first word of a surname-fi… — plural (#766)	b	—
t	nl	article	#18 (gist)	never addressed; 't ships, the apostrophe-less form is a bare letter	c	DF
साहिब	pa	honorific/attached	vocab/pa.toml:25	जी ends बनर्जी, मुखर्जी, चटर्जी, so a glued peel would strand a bare virama (बनर् + जी); the 殿 criterion in a non-CJK script (decisions/indic-honorifics) — the clash is inside the word's own script and language	b	—
i	pl	conjunction/may_be_initial	decisions/3-0-reevaluations	as ca i	a	pack removes a property
pan	pl	salutation	vocab/pl.toml:4	Polish and Czech Mr., and the Chinese surname Pan, which leads — clash is with the Chinese surname Pan	a	DF+ST
e	pt	conjunction/may_be_initial	PR #132; decisions/3-0-reevaluations; decisions/P8	stays marked in the default; "a locale pack cannot express 'remove e'"	a	pack removes a property
senhoro	pt	salutation	PR #99	not a standard Portuguese form	b	—
dr.ª	pt	salutation	#840; #606	never addressed	c	—
prof.ª	pt	salutation	#840; #606	never addressed	c	—
professora	pt	academic	#606; #804	#804 open	c	—
terceiro	pt	generation	#843	never decided	c	DF
f	pt-BR	agnomen	#843 (derek73 2026-10-25); decisions/S11	written as initials; "would be default = false, a phase 3 pack question"	a	DF
fil	pt-BR	agnomen	vocab/pt-BR.toml:4	Filho written short in 10 tse-donors filed/registry pairs (2026-10-25), but a Slavic and Polish surname (US Census 220, Wikidata 6 with Poland and Ukraine) that a name… — clash is the Slavic/Polish surname Fil; strict pt-BR drops it	a	DF+ST
n	pt-BR	agnomen	#843 (derek73 2026-10-25); decisions/S11	written as initials; "would be default = false, a phase 3 pack question"	a	DF
sb	pt-BR	agnomen	vocab/pt-BR.toml:38	Sobrinho written short in 11 tse-donors filed/registry pairs (2026-10-25), but also S.B., the Bachelor of Science, written SB by 23 ClinicalTrials.gov officials (9 aft… — clash is the en credential SB; prefer pt-BR settles it only if the agnomen is a (default = false) vocabulary reading	a	DF+PR
fi	pt-BR	agnomen	#843	never addressed (abbreviations and misspellings)	c	DF
fiho	pt-BR	agnomen	#843	never addressed (abbreviations and misspellings)	c	DF
fiilho	pt-BR	agnomen	#843	never addressed (abbreviations and misspellings)	c	DF
fl	pt-BR	agnomen	#843	never addressed (abbreviations and misspellings)	c	DF
ne	pt-BR	agnomen	#843	never addressed (abbreviations and misspellings)	c	DF
и	ru	conjunction/may_be_initial	vocab/ru.toml:60	#267 blessed the Cyrillic joining, and marking it would break that reading in the one-case spellings where nothing else decides — #267 Cyrillic joining	b	—
ст	ru	generation	vocab/ru.toml:46	barely attested after a name, and a plausible false positive among two-letter Cyrillic abbreviations; deferred since #269 — barely attested; Russian data decides	c	DF
ml	sk	generation	vocab/sk.toml:4	pack-shaped: real and standard, but collides with ML initials in other languages written in Latin script — as cs ml	a	DF
st	sk	generation	vocab/sk.toml:18	pack-shaped: real and standard, but St. is Saint in other languages written in Latin script — cs st is already default = false; these three are excluded instead	a	DF
ml	sl	generation	vocab/sl.toml:4	pack-shaped: real and standard, but collides with ML initials in other languages written in Latin script — as cs ml	a	DF
st	sl	generation	vocab/sl.toml:18	pack-shaped: real and standard, but St. is Saint in other languages written in Latin script — cs st is already default = false; these three are excluded instead	a	DF
fru	sv	salutation/after_given	vocab/sv.toml:4	a Cameroonian surname (~10,800, the note above), and middle names are drawn from surnames — surname evidence is Cameroonian, another language	a	DF-prop+ST
norr	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
ofver	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
soder	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
upp	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
vast	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
vaster	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
väst	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
väster	sv	particle	PR #132	never judged: #132 was closed for #182's last_base view; as the nb farm-name words	c	DF+ST
thiru	ta	salutation	vocab/ta.toml:4	Tamil Mr., and a Tamil given name — same-language given name	c	—
นาง	th	salutation	#317 (open)	needs a head-peel mechanism and a Thai reader's vetting	c	rule (attached head peel)
นางสาว	th	salutation	#317 (open)	needs a head-peel mechanism and a Thai reader's vetting	c	rule (attached head peel)
นาย	th	salutation	#317 (open)	needs a head-peel mechanism and a Thai reader's vetting	c	rule (attached head peel)
bay	tr	salutation	vocab/tr.toml:4	Turkish Mr., and a surname borne in Hungary, which writes the surname first, where the title claim acts, so it fails C-i — clash is the Hungarian surname, surname-first; strict tr drops it	a	DF+ST
baba	tr	generation	vocab/tr.toml:19	collides inside Turkish, and Turkish puts the junior word before the name — collides inside Turkish, and is written before the name	b	—
kıdemli	tr	generation	vocab/tr.toml:25	a rank word (Kıdemli onbaşı); Turkish writes Jr. or a parenthesis, (oğul) — rank word	b	—
й	uk	conjunction/may_be_initial	vocab/uk.toml:49	#267 blessed the Cyrillic joining, and marking it would break that reading in the one-case spellings where nothing else decides — #267 Cyrillic joining	b	—
і	uk	conjunction/may_be_initial	vocab/uk.toml:41	#267 blessed the Cyrillic joining, and marking it would break that reading in the one-case spellings where nothing else decides — #267 Cyrillic joining	b	—
пан	uk	salutation/after_given	vocab/uk.toml:4	a surname in Cyrillic (Wikidata), and middle names are drawn from surnames — surname/given evidence is mostly US Census bearers of the same word; a property cannot be default = false today	c	DF-prop
bon	und	particle	vocab/und.toml:13	no particle use found in any name, and it is borne as a given name and as a surname (the counts on its name readings), which the particle reading joined to the next wo… — no particle use found (#766)	b	—
dí	und	particle	vocab/und.toml:19	no particle use found in any name; its Wikidata bearers are Chinese given names romanized Dí, which the particle reading joined to the next word — no particle use found (#766)	b	—
janab	ur	salutation	vocab/ur.toml:4	unverified, so not shipped: the Devanagari जनाब and Bengali জনাব ship, and the Latin spelling needs its own check (the sri/shri pattern) — unverified; native spellings ship	c	—
ông	vi	salutation	vocab/vi.toml:8	Ông is a Vietnamese surname and Vietnamese names lead with the surname, the position the title claim acts on; pinned by the case row salutation_that_leads_a_surname_st… — same-language surname in the leading slot; no pack settles it	b	—
bà	vi	salutation	vocab/vi.toml:4	Vietnamese Mrs., refused as ông's pair without evidence of its own: #606 judged the excluded side from general knowledge, with no count taken (decisions/salutation-tit… — unmeasured; ông's pair	c	—
ajala	yo	generation	vocab/yo.toml:12	a Yoruba surname — same-language surname	b	—
东野	zh	surname (ZH pack)	decisions/W1 Declined	no attested bearer, or the first character alone opens names	b	—
东门	zh	surname (ZH pack)	decisions/W1 Declined	no attested bearer, or the first character alone opens names	b	—
冯	zh	preposition	#671; decisions/W2	a top-30 surname	b	—
南门	zh	surname (ZH pack)	decisions/W1 Declined	no attested bearer, or the first character alone opens names	b	—
君	zh	honorific	vocab/zh.toml:39	王君 is a complete Chinese name, 君 being a common given-name final; its kana spelling くん ships — same-language given-name final	b	—
宗政	zh	surname (ZH pack)	decisions/W1 Declined	no attested bearer, or the first character alone opens names	b	—
小	zh	generation	#671; decisions/W2	written before the name; a common given-name character	b	—
張李	zh	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	married double surnames: longest match would cut 李林伟	b	—
張陳	zh	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	married double surnames: longest match would cut 李林伟	b	—
归海	zh	surname (ZH pack)	decisions/W1 Declined	no attested bearer, or the first character alone opens names	b	—
成公	zh	surname (ZH pack)	decisions/W1 Declined	no attested bearer, or the first character alone opens names	b	—
李林	zh	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	married double surnames: longest match would cut 李林伟	b	—
王李	zh	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	married double surnames: longest match would cut 李林伟	b	—
谷梁	zh	surname (ZH pack)	decisions/W1 Declined	no attested bearer, or the first character alone opens names	b	—
部長	zh	professional_honorific/attached	vocab/zh.toml:140	長 alone is a borne given name, so 阿部長 can be 阿部 + 長 (#707). The measurement found no such bearer, as it found none for the words that shipped; the refusal rests on the… — the clash is inside the word's own script and language	b	—
部长	zh	professional_honorific/attached	vocab/zh.toml:148	the simplified twin of 部長, declined with it (#707) — the clash is inside the word's own script and language	b	—
陳李	zh	surname (ZH pack)	decisions/vocabulary-exclusions (#695)	married double surnames: longest match would cut 李林伟	b	—
馮	zh	preposition	#671; decisions/W2	a top-30 surname	b	—
博士	zh	credential/attached	vocab/zh.toml:14	glued 田中博士 IS Tanaka Hiroshi, an attested given name — the clash is Japanese (田中博士 = Hiroshi); strict zh would drop it, but Han is shared by ja and zh	c	DF-prop+ST
isizukulwane	zu	generation	vocab/zu.toml:4	CLDR label translation: it means 'generation' — means "generation"; a CLDR label	b	—

Phase 3 probe: missing foreign titles and function-word/credential collisions

Read-only research for #753 (vocabulary Phase 3, language packs). Measured 2026-10-27 on cfe8bf92 (master after 2.7.0, 2.8 cycle open). No tracked file changed. Local corpora were read from the cache; nothing here quotes a real name. Every name below is made up.

Words already recorded as excluded (pan, ing, heer, mst, the cs/pl salutations) belong to the kept-out sweep in the comments above. They are marked "(recorded)" here and not reargued.

Summary

The honorifics are covered. Occupational titles in other languages are not. Of the 9 "Mr/Mrs" families checked (es, pt, fr, de, it, nl, pl, cs, tr), the vocabulary holds almost every bare form: Señor, Senhora, Monsieur, Herr, Frau, Signora, Mevrouw, Paní, Slečna. The missing ones are Don, Doña and Dona, Pan and Pani, the Turkish forms (Bey, Hanım, Bayan, Sayın), Heer, and French M..
Across the eight occupational families (professor, doctor, engineer, judge, senator, reverend, lawyer, architect), the vocabulary holds the English word, plus prof, dr, dra, mgr, pastor and imam. It holds no bare full-word form in any other language: 0 of about 120 forms checked.
The abbreviations already work by shape. Every period abbreviation tried reads as a title through rules/H2's leading shape, including the interior-period and stacked forms: Profa., Prof.ª, Dott.ssa, Sig.ra, Dipl.-Ing., Dr hab. inż., MUDr., Doç. Dr., Mr. drs.. The whole gap is bare words, which become the given name (Profesor Juan Pérez → given 'Profesor').
No corpus writes these words often enough to rank them.
- The PSC title box held 888,830 filed titles. Only 1,291 (0.15%) are words outside titles ∪ titles_ambiguous. Most of those are English typos (mt, nr, mis, mss), English phrases (the hon, executors of, managing director) or job titles.
- Foreign filers on a UK form type English titles: Romanian, Polish, Italian, French and Spanish filers each write "mr" in 60–70% of their boxes.
- The largest foreign title words in the PSC misses are m (French M.: 70 records, plus 40 non-UK filers' m boxes), engr/eng/ing, dott, sig, drs, mudr, jonkheer, pan and bay, each at 2–11.
- Across the 8 local corpora, the missing bare occupational words almost never open a name as a title (tools/vocab/word_report.py --rank). Where they open a name at all, the word is a surname or given name: richter 327, pan 270, hakim 186, don 1,045, dona 59.
- The committed corpora (6,853 names) hold none, apart from CLDR/rules rows for sig, pani, sigra and dipl-ing.
- So no language's pack can be justified by corpus incidence. It would have to rest on primary sources (style guides, government forms), and on the structural problem in (4).
The structural problem is bigger than the wordlist. Adding the titles is not enough:
- With profesora and ministro added, Profesora de Historia Ana López still reads as title 'Profesora' and family 'de Historia Ana López'. Rule P1 folds the never-given particle de, and rule P4 says titles don't move the opening position.
- Ministro da Fazenda Carlos Souza reads as given 'da', because da is an ambiguous particle.
- Making de and da conjunctions globally fixes both phrases (title 'Profesora de Historia'), but it breaks every surname: Ana de Souza → given 'Ana de Souza'.
- What a pack needs is a connective that counts only inside a title run. That is an H9/P3 rule question, not a wordlist one.
Function-word/credential collisions: 9 distinct words (15 language/word pairs).
- Three are already particles or conjunctions: do, i, abd.
- Six are not: am (de), ed (it), od (it, pl, cs), v (cs), se (cs), ala (ar).
- The harm stays narrow. A credential reading needs the word after a full name and a comma, or a trailing bare word with the count to spare. In the middle of a name each one reads as a middle word.
- One made-up shape loses its given name outright: Haddad, Abd reads as family 'Haddad', suffix 'Abd'.
- Particle words that a pack would add (el, des, les, il, lo, zum, zur, ten, ben, bou, ul, abi, dalla) collide with no suffix set.

Languages where enough is missing to justify a pack (bare occupational forms, plus honorifics the default lacks):

es and pt. The whole occupational family is missing, along with Don/Doña/Dona, and they have the de/da phrase problem.
fr. Maître, Docteur, Professeur and M. are missing.
it. Dottore and Avvocato are missing, though the abbreviations work by shape.
tr. The postnominal Bey and Hanım are a separate shape (see shape 7).
de, pl and cs. Here the bare forms rarely precede a name in writing, and the abbreviations dominate and already work.

1. Titles that real filers write and the vocabulary misses (psc-individuals)

Population: 1,040,708 PSC records. 888,830 have a filed title box. The fold is local_corpora.word_key. A box hits when every word in it is in titles ∪ titles_ambiguous.

shape of the missed box	records	share of filed titles
bare word	641	0.07%
multi-word box	533	0.06%
other dotted (`M.`, `Dipl-Ing`, `H.E.`)	86	0.01%
period abbreviation (`Engr.`, `Ing.`, `Brig.`, `MUDr.`, `Drs.`)	31	<0.01%
total	1,291 records, 609 distinct boxes	0.15%

Top misses, as boxes in key form, with each box's top nationalities. Boxes that were a filer's own name are left out.

box	n	shape	nationalities
m (M.)	70	dotted	british 29, french 25, moroccan 2
mt	48	bare (typo)	british 37
managing director	42	multi	british 23, german 3, american 3
md	39	bare	bangladeshi 24, british 12 (a given name, #853)
mis	29	bare (typo)	british 14, romanian 3
owner	27	bare	american 8, british 5
nr	25	bare (typo)	british 21
the estate of / executors of / exors of (all variants)	~110	multi	british
the hon / the honourable / the rt hon / the most hon	~50	multi	british (these PARSE fine: title 'The Hon')
my, mss, m/s, mra, mrd, mrr, misd, …	~80	bare (typos)	british
revd	15	bare	british 14
engr	11	period	british 4, pakistani 4, nigerian 1
eng	7	bare	egyptian 3, romanian 2, iranian 1
ing	7	period	czech 4, slovak 2, austrian 1 (recorded)
mst	6	bare	bangladeshi 6 (recorded; forename box 43 of 49, a given-name abbreviation like md)
directeur	4	bare	comoran, french, tunisian
hrh prince / hrh / hh / hh sheikh / hsh prince	13	bare/multi	saudi, emirati, qatari, liechtenstein
sig	3	bare	italian 3
alhaji	2	bare	nigerian, british
drs	2	period	dutch 2
dipl-ing / dipl ing	4	dotted/multi	austrian, dutch, german, slovak
dott	2	bare	italian, ghanaian
mudr	2	period	slovak, czech
jonkheer	2	bare	dutch 2
pan	2	bare	polish 2 (recorded)
bay	2	bare	turkish 2
very reverend / rt revd / revd canon	6	multi	british, irish

Title boxes from non-UK nationalities, excluding the English core (mr, mrs, ms, miss, dr, prof, sir, lady, mx, dame, lord, rev). Hits and misses mixed:

director 247, ceo 99, madam 69, mister 62, professor 61, me 50, sr 42, m 40, doctor 31, md 26, owner 22, president 19, reverend 14, mis 13, founder 13, herr 13, sheikh 12, mme 11, master 11, monsieur 10, sra 8, baron 8, domnul 8, engr 7, ing 7, frau 4, pastor 4, directeur 4, signor 3, sig 3, madame 3, …
Native honorifics stay in single digits, beside thousands of "mr" from the same nationalities.

The English words this turned up in passing, with nothing recorded about them (no TOML entry, no decisions mention):

revd, hrh, hh, hsh, eur (Eur Ing), engr, jonkheer, alhaji.
very as a modifier of Reverend: Very Reverend John Smith → given 'Very'.
HRH Prince Ali Haddad → given 'HRH'. The H9 phrase never starts, because HRH is not a title.
The Executors of John Smith → given 'The Executors of John'. This is garbage input, but 110 filers wrote it.

UKEC (15,213 multi-word names): 8,700 lead with a listed title. Among leading words not listed, only the (29, which parses fine as "The Hon"), m (9), don (3) and md (3) touch the candidate list. One period-shaped miss. UKEC adds nothing for foreign titles.

2. Translated title families

Key: T = listed title. shape = not listed, but a title by rules/H2 when it opens a name with its period. — = missing (a bare form reads as the given name). risky = borne as a name (counts from tools/name_usage.py --no-wikidata for the six words checked, or from the PSC forename box). seen = the incidence the local corpora show where the word opens a name or follows a comma (word_report --rank). Nearly all of those openings are surnames or given names, not titles.

family	es	pt	fr	de	it	nl	pl	cs	tr	ar-Latin
professor	profesor —, profesora —, catedrático —, Prof. T, Profa. shape	professor T, professora —, Prof.ª shape	professeur —, professeure —, Pr. shape	Professor T, Professorin —, Prof. T	professore —, professoressa —, Prof. T	hoogleraar —, prof. T	profesor —, prof. T	profesor —, profesorka —	profesör —, Doç. shape, Doçent —	ustaz — / ustadh —
doctor	doctor T, doctora —, Dra. T	doutor —, doutora —, Dr.ª shape	docteur —, docteure —, Dr T	Doktor —, Doktorin —, Dr. T	dottore —, dottoressa —, Dott. shape, Dott.ssa shape	dokter —, drs. shape (2 PSC)	doktor —	MUDr./JUDr./PhDr./RNDr. shape (seen: phdr 10, rndr 6, mudr 26 after a comma)	doktor —, hekim —	duktur/doktur —
engineer	ingeniero —, ingeniera —, Ing. shape	engenheiro —, engenheira —, Eng. shape	ingénieur —	Ingenieur —, Dipl.-Ing. shape, Dr.-Ing. shape	ingegnere —, Ing. shape	ir. shape, Ir —	inżynier —, inż. shape	Ing. shape (bare ing recorded)	mühendis —, Müh. shape	muhandis —, Engr. shape (11 PSC), Eng — (seen 109 opens, 2 forename vs 7 title box)
judge	juez —, jueza —, magistrado —	juiz —, juíza —, desembargador —	juge —	Richter — risky (US surname 29,880; 327 corpus opens)	giudice —	rechter —, mr. T	sędzia —	soudce —	yargıç —, kadı —	qadi —, qazi — (48 opens, surname), hakim — risky (US surname 3,165, SSA given 2,361+)
senator	senador —, senadora —	senador/a —	sénateur —, sénatrice —	Senator T	senatore — (14 opens), senatrice —	senator T	senator T	senátor —	senatör —	—
reverend/pastor	reverendo —, pastora —, padre — (37 opens), párroco —, cura —	padre —, pastora —, frei — risky (US surname 2,419; 39 opens)	père —, pasteur —, abbé —, révérend —	Pfarrer —, Pastor T	parroco —, don —	dominee —, ds. shape	ksiądz —, ks. shape	farář —	papaz —, imam T, hoca —	imam T, imaam —, sheikh Tamb, shaykh T
lawyer	abogado —, abogada —, Lic. shape (seen 20 after a comma), licenciado —	advogado —, advogada —	maître —, Me T, avocat —	Rechtsanwalt —, RA T+SA	avvocato —, Avv. shape	advocaat —, mr. T	adwokat —, mecenas —, mec. shape	advokát —, JUDr. shape	avukat —, Av. shape (av is a P!)	muhami —
architect	arquitecto/a —, Arq. shape	arquiteto/a —, Arq. shape	architecte —	Architekt —, Architektin —	architetto/a —, Arch. shape	ir. shape	architekt —	architekt —	mimar —	—
honorific	Señor/a T, Srta T, Don — risky (SSA given 168,839+; 1,045 opens, 2,820 after a comma), Doña —	Senhor/a T, Dona — risky (SSA 19,032+, 158 after a comma)	Monsieur T, Madame T, Mme T, M. — (a bare initial: 1,604 opens, 74 PSC title boxes)	Herr T, Frau T	Signor/a T, Sig. shape, Sig.ra shape	Mevrouw T, dhr T, heer (recorded), mw —	Pan (recorded), Pani — (cs paní T), Panna —	Pán —, Paní T, Slečna T	Bay — risky (122 opens), Bayan —, Bey — postnominal, risky (US surname 5,453), Hanım — postnominal, Sayın —	Alhaji — (151 opens, title box 2 vs forename box 1)

Notes:

pr (French Pr., professeur) is listed as missing above. It already appears 1,366 times after a comma, and those PR/Pr are credentials (Public Relations, etc.).
av (Turkish Av., Avukat) is a nb/sv particle. Av. Ahmet Yılmaz still reads title 'Av.', because the shape rule runs first.
don and dona are the clearest "risky" words. The Spanish honorific Don is also a common English given name, and Dona is one too. A pack could list them only for a caller who opted in.

3. Function words that collide with credentials

Checked: the articles, prepositions and conjunctions of es, pt, fr, de, it, nl, pl, cs and tr, plus Arabic-in-Latin particles (≈250 words). Each was checked against suffix_acronyms (SA), suffix_acronyms_ambiguous (SAamb), suffix_words (SW) and suffix_words_ambiguous, and against particles (P), conjunctions (C), conjunctions_ambiguous and bound_given_names.

word	languages	in sets	already P/C?	what `parse()` gives on a made-up name
do	pt, pl, cs	SA, SAamb, P	yes	`Maria do Carmo Velasco` → given 'Maria', family 'do Carmo Velasco'; `Ana Velasco, DO` → suffix 'DO'; `Ana Velasco Do` → family 'Do' [SUFFIX_OR_NAME]
am	de	SA	no	`Klaus am Rhyn` → middle 'am', family 'Rhyn' (family should be 'am Rhyn'); `Klaus Rhyn, AM` → suffix 'AM'
i	it (pl.), pl, cs, ca	SW, C, Camb	yes	`Joan Garcia i Puig` → family 'Garcia i Puig' (good); `Joan Garcia I` → suffix 'I' [SUFFIX_OR_NAME]
ed	it	SA, SAamb	no	`Paolo Rossi ed Bianchi` → middle 'Rossi ed'; `Bianchi, Ed` → given 'Ed' [SUFFIX_OR_NAME] (the gate holds)
od	it (archaic), pl, cs	SA	no	`Anna od Krzyża` → middle 'od' (a religious-name shape: should be family or a byname 'od Krzyża'); `Jan Lesny, OD` → suffix 'OD'
v	cs	SW (roman numeral)	no	`Jan Novak v Lese` → middle 'Novak v'; `Jan Novak, V` → suffix 'V' (intended)
se	cs	SA, SAamb	no	no realistic name shape; a reflexive, not a name word
ala	ar	SA, SAamb	no (a given name)	`Ala Haddad` → given 'Ala'; `Haddad, Ala` → given 'Ala' [SUFFIX_OR_NAME] (the gate holds)
abd	ar	SA, bound	yes (bound)	`Abd Rahman Haddad` → given 'Abd Rahman' (good); `Haddad, Abd` → suffix 'Abd', no given name

Count: 9 distinct words, 15 language pairs. Six of the words are not in any particle or conjunction set. The real harm is narrow:

Haddad, Abd, where abd is an unambiguous acronym and loses a given name.
am and od, where the particle is not joined to the family name. That one is a missing particle, not a credential problem. Once listed as particles, the credential reading after a comma would remain, as it does for do.

A pack that adds am or od as a particle meets the credential set exactly as do already does. So the per-pack overlap check has a precedent: do sits in P ∩ SAamb today.

4. Example shapes for the design session

All made up. "Today" is parse() on cfe8bf92. "Wanted" is what a pack would deliver.

#	shape	today	wanted (with the pack)	what decides it
1	`Profesora de Historia Ana López`	given 'Profesora', family 'de Historia Ana López'	title 'Profesora de Historia', given 'Ana', family 'López'	Adding `profesora` gives title 'Profesora', family 'de Historia Ana López' (the P1 fold). Making `de` a conjunction globally fixes this one and breaks `Ana de Souza` (→ given 'Ana de Souza'). So it needs `de` as a connective inside a title run only.
2	`Ministro da Fazenda Carlos Souza`	given 'Ministro', family 'da Fazenda Carlos Souza'	title 'Ministro da Fazenda', given 'Carlos', family 'Souza'	With `ministro` added, it reads given 'da', because `da` is an ambiguous particle and the title leaves it opening the name. This is worse than today.
3	`Senador Juan Pérez`, `Juez …`, `Doutor …`, `Maître …`, `Rechtsanwalt …`	given = the title word	title = the word	A plain wordlist; adding each one works (verified for senador, padre, juez with a custom Lexicon).
4	`Directeur Général Jean Dupont`	given 'Directeur', middle 'Général Jean'	title 'Directeur Général'	A modifier after its head: the H9 "qualifier written after its head" shape, now in French. Works once both are titles.
5	`Don Juan Pérez` / `Dona Maria Souza`	given 'Don' / 'Dona'	title, but only for a Spanish/Portuguese caller	The pack is opt-in because the default can't take the word: Don is an English given name (SSA 168,839+).
6	`M. Jean Dupont`	given 'M.', middle 'Jean'	title 'M.' under a French pack	A single letter is never H2's shape (`{2,}`), so this is an initial-vs-title fork that only a language decides. 70 PSC boxes.
7	`Ahmet Yılmaz Bey`, `Ayşe Yılmaz Hanım`	family 'Bey' / 'Hanım'	title (postnominal honorific)	A trailing bare title. The doctrine keeps bare trailing words as names (H5); H8's after-given salutation is family-comma only. Turkish puts the honorific after the given name with no comma (`Ahmet Bey`), so this is a new position rule. Bey is also a US surname (5,453).
8	`Juan Pérez, Profesor`	given 'Profesor', family 'Juan Pérez' (the comma reads as a family comma)	title 'Profesor'	With the word added, this reads correctly (title 'Profesor'), so it is the vocabulary that decides.
9	`Haddad, Abd`	family 'Haddad', suffix 'Abd'	given 'Abd'	A function word read as a credential: `abd` is in SA unambiguously. Move it to SAamb, or let a pack's given/bound reading derive the ambiguity.
10	`Klaus am Rhyn` / `Anna od Krzyża`	middle 'am' / 'od'	family 'am Rhyn' / 'od Krzyża'	Particles missing from the default. Adding them meets SA (`am`, `od`), with `do` as the precedent.
11	`HRH Prince Ali Haddad`, `HH Sheikh …`	given 'HRH' / 'HH'	title 'HRH Prince'	English and Gulf style abbreviations, not a pack question. Listed so it doesn't get lost.
12	`Mst Rahima Khatun`	given 'Mst'	given 'Mst' (correct today)	PSC: 43 of 49 filers put `Mst` in the forename box. If a bn pack ever lists মোসাঃ's transliteration, it is a given-name abbreviation like `Md` (#853), not a title.

Shapes that already work and need nothing (with their periods): Dott.ssa, Sig.ra, Prof.ª, Profa., Dra., Dipl.-Ing., Prof. Dr.-Ing., Dr hab. inż., Mr. drs., Ir., MUDr., Doç. Dr., Av., Ks., Mec., Lic., Arq., Engr., Eng.; Ing. Jan Novák, CSc. (suffix 'CSc.'). Herr Prof. Dr. Hans Müller reads title 'Herr Prof. Dr.'. Hans Müller, Dipl.-Ing. reads title 'Dipl.-Ing.' after the comma.

5. Questions for the Phase 3 design

A connective inside a title run. Should a pack declare its title-phrase connectives (de, da, do, del, für, voor), as words that join a title run (H9/P3) while staying particles everywhere else? Shapes 1 and 2 show that no wordlist edit can deliver this. Is it a pack field, or a property on the particle's reading, scoped to "after a title"?
What justifies a pack word when no corpus writes it? Every corpus we hold is a registry or catalog that strips titles, or a UK form where foreigners type "Mr". Is a primary source (a style guide or a government form) enough? Or does Phase 3 need a corpus that writes titles in other languages before shipping one, such as news bylines or JRC's multilingual spellings scanned for title-opening runs?
Gendered pairs. Most missing families come as m/f pairs (profesor/profesora, doutor/doutora, ingeniero/ingeniera). Does the pack model gender as salutation's gender does for occupational titles, and does anything read it?
Risky honorifics as opt-in only. don, dona, bay, bey, pan, M., hakim and richter cannot be default titles. Is "a pack word may be one the default refuses" the core licence of a pack? If so, does a pack word override a default excluded entry, and how is that override recorded?
Postnominal honorifics. Turkish Bey/Hanım, and Japanese 様 (already handled for CJK by W2's glue): does a pack get to declare an after-given salutation outside the family-comma position H8 allows?
Credential overlap per pack. A pack adding am or od as a particle lands in SA, as do already does. Should the build's derived ambiguity (phase 2's "a name reading beside a credential makes it ambiguous") also fire for a particle reading, so that Haddad, Abd's class is closed by construction rather than by a list?
Single-letter titles. French M. is the one honorific that rules/H2's {2,} shape can never reach, and that loses to the initial reading. Is a one-letter title a pack-only exception to the initial veto?
The English stragglers found here (revd, hrh, hh, hsh, very, eur, engr, jonkheer, alhaji): are they default-vocabulary work for Should Baron, Bishop, Judge and the other title words borne as names read as names outside a title phrase? (Baron Davis → title 'Baron', no given name) #913's batches rather than Phase 3? None is recorded anywhere.

Navigation Menu

Repository navigation

Phase 3: language packs that prefer or restrict their language #753

Phase 3 input: words kept out of the default vocabulary (#753)

The two figures

What the inventory says

Languages with enough candidates to justify a pack

The locale packs today

Summary

Questions for the Phase 3 design

Method and limits

Full table, part 1 of 2

Full table, part 2 of 2

Phase 3 probe: missing foreign titles and function-word/credential collisions

Summary

1. Titles that real filers write and the vocabulary misses (psc-individuals)

2. Translated title families

3. Function words that collide with credentials

4. Example shapes for the design session

5. Questions for the Phase 3 design

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Phase 3: language packs that prefer or restrict their language #753

Description

Activity

derek73 commented on Oct 24, 2026

derek73 commented on Oct 25, 2026

derek73 commented on Oct 28, 2026

Phase 3 input: words kept out of the default vocabulary (#753)

The two figures

What the inventory says

Languages with enough candidates to justify a pack

The locale packs today

Summary

Questions for the Phase 3 design

Method and limits

derek73 commented on Oct 28, 2026

Full table, part 1 of 2

derek73 commented on Oct 28, 2026

Full table, part 2 of 2

derek73 commented on Oct 28, 2026

Phase 3 probe: missing foreign titles and function-word/credential collisions

Summary

1. Titles that real filers write and the vocabulary misses (psc-individuals)

2. Translated title families

3. Function words that collide with credentials

4. Example shapes for the design session

5. Questions for the Phase 3 design

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions