What does an initial look like in Devanagari, Bengali and the scripts like them? (एम. एस. वर्मा gives initials() 'ए. ए. व.') #1273

Activity

1. derek73

   self-assigned this

   on Nov 9, 2026
2. derek73

   added

   question

   on Nov 9, 2026
3. derek73

   added a parent issue on Nov 9, 2026

   • Which scripts have case, and which write initials? (ჯეკ MA, ကြည် ကြည် ဌေး, एम. एस. वर्मा) #1256
4. derek73

   added

   initialsReading initials and the initials() view (rules/T)Reading initials and the initials() view (rules/T)

   non-latin-scriptNames written in a script other than Latin (rules/W)Names written in a script other than Latin (rules/W)

   and removed

   question

   on Nov 10, 2026
5. derek73

   added a commit that references this issue on Nov 10, 2026

   Count names in Georgian, Armenian, Thai, Ge'ez, Burmese and Gurmukhi …

   ...

   d95a5ca
6. derek73 commented on Nov 12, 2026

   

   derek73

   on Nov 12, 2026

   OwnerAuthor

   More actions

   Data points from the #1254 work (PRs #1347 and #1348), which made two rules depend on what T5 tags as an initial. Readings and counts re-taken on master 670b349, 2026-11-12, unless marked.

   Where the line falls today. T5 tags a Devanagari word as an initial when it is one bare letter and a full stop. A letter with a vowel sign is two code points and gets no tag, with or without the full stop.

   Input	Tags	Default	Policy(initials_as_family=True) / parser_for(locales.NE)
   झलक बहादुर आ. म.	आ. [initial], म. [initial]	middle बहादुर आ., family म.	middle बहादुर, family आ. म. (moves, rules/O8)
   राम बहादुर बि. क.	क. [initial] only	middle बहादुर बि., family क.	the same (no run: बि. is no initial)
   राम बहादुर वि क	none	middle बहादुर वि, family क	the same
   राम बहादुर बि.क.	none	middle बहादुर, family बि.क.	the same (one word, the family by position)
   राम क.	क. [initial]	given राम, family क., reports order	no report
   राम बि.क.	none	given राम, family बि.क.	no report either way
   Ram Bahadur B K	B, K [initial]	middle Bahadur B, family K	middle Bahadur, family B K
   Ram B.K.	shape:acronym	family B.K., reports order, suffix-or-name	reports suffix-or-name

   So the same person's name moves in Latin and stays in Devanagari, and the romanized byline reports where the Devanagari one is silent. rules/O8's Accepted clause and rules/O4's third Except bullet state both as limits and point here.

   Counts in nepal-candidates (237,408 distinct corpus spellings: 119,285 holding a Devanagari letter, 118,123 not; a script reading corpus.jsonl, each name parsed under parser_for(locales.NE) and under the same parser with initials_as_family off):

   • 551 romanized spellings move under the pack's switch; 1 Devanagari spelling moves (the आ. म. shape above).
   • 598 Devanagari spellings end in two or more spaced one-letter words (one letter with its vowel signs, a full stop allowed) after a longer word: 597 keep their reading, 1 moves. The commonest runs: वि क 259, बि क 109, वि. क. 59, के सी 55, के. सी. 23, बि. क. 17. This matches the 597 that rules/O8 quotes from 2026-11-11.
   • 2,410 more end in the same letters written as one dotted word (वि.क. 918, वि.क 383, के.सी. 311, बि.क. 270, के.सी 137, बि.क 113): none moves, and none needs to for the family name, which position already gives.
   • Order reports (default parser): 1,580 Latin spellings report order; 0 Devanagari spellings do. Of Devanagari spellings shaped one longer word followed by nothing but such letters (spaced or as one dotted word), I count 1,891, all silent: 1,381 written with a full stop somewhere, 510 bare. decisions/O4's Report the order on a name of one word followed by initials (#1254) #1348 entry gives 1,407 for "names of the shape" from that PR's review (2026-11-11), a narrower cut whose recipe is not in the tree; treat it as a dated snapshot.

   What this adds to the question.

   1. The surname abbreviations (बि.क. for B.K., के.सी. for K.C.) are spelled with a vowel sign because they transliterate the Latin letter's name (बि = "bi", के = "ke"). So in this population the spelled-letter case of this issue (एम. for M.) and the vowel-sign case are the same thing: a syllable standing for a Latin letter. A rule that takes only bare letters as initials catches 1 name in 598.
   2. A shape test of "one consonant or vowel with its vowel signs, then a full stop" would tag बि., के., सी. and also every one-syllable word written with a full stop. The bare spaced forms (वि क, 259) have no full stop at all, so a full-stop test alone still leaves most of the 598 untouched.
   3. A closed list of the syllables that spell Latin letters, per script, is the other route this issue names; वि/बि, क, के, सी, जी cover nearly all of the runs counted here.
   4. Whatever is chosen is read by O8 (the pack's switch), O4's order report, T7, and initials(), so राम बहादुर वि क under locales.NE and the report on राम बि.क. are the two rows to put in the grid.

   🤖 Generated with Claude Code