Repository navigation
Should د. and the Iraqi academic abbreviations read as titles? (د. أحمد علي حسن gives given 'د.') #1149
Description
parse("د. أحمد علي حسن") gives given 'د.', middle 'أحمد علي', family 'حسن'. د. is the Arabic abbreviation for "Dr." The Iraqi academic forms aren't listed either:
أ.د.(professor doctor):أ.د. أحمد علي حسنgives given 'أ.د.'م.د.(lecturer doctor):م.د. أحمد عليgives given 'م.د.'
An Iraqi polytechnic's 2023–24 staff list (https://iku.atu.edu.iq/?page_id=160) writes د. on 3 of its 125 names, both spaced (د. X) and glued (د.X).
Why د. is refused today (decisions/vocabulary-exclusions, ar.toml's excluded:academic entry): the edge-period fold leaves a bare د. A single letter matches an initial, so the title would swallow initials, as with Greek κ. But Arabic script does not seem to write initials:
- 0 single-letter-plus-period words in JRC-Names' 17,080 Arabic-script spellings
- 0 in
corpus_wikidata_native.jsonl's 212
The Bengali ড already ships on the same reasoning and pays the known cost (decisions/indic-honorifics).
Questions:
- Should
د.ship the wayডdoes, with its cost (د. حسنreads an empty given)? - Should
أ.د.,م.د.andم.م.ship? Each has more than one letter, so the fold leaves no bare letter and they avoid the initials trap. - The glued form
د.أحمدis a separate shape and not Arabic-only:Dr.John Smithgives title 'Dr.John'. Is that in scope here or its own issue?
What to measure first: a source that writes Arabic titles. JRC-Names is harvested from news text and holds none (its 5 names starting with a bare د are cut-off spellings of محمد). University staff lists are the obvious candidate.
Activity
derek73 commentedon Nov 8, 2026
The King Saud University faculty directory, where each faculty member writes their own entry, measures what this issue asked for (2026-11-07, every 10th page, 401 names in Arabic script only):
د.: 52 spaced (د. X), 9 glued (د.X) and 2 bare (د X, with no period)أ.د.أ.(Ustadh)None of these reads as a title today; only the spelled-out
الدكتور,دكتورandالأستاذ الدكتورdo. The directory holds no Arabic-script initials, consistent with the 0 found in JRC-Names.Egyptian and some Saudi entries write the title with a slash:
أ.د/,أ.م.د/,د/,م.م/,م/,دكتورة/,الدكتور/(Kafr El-Sheikh University's faculty guide, KSU).د/ منى أحمد سالمgives given 'د/', andDr/ John Smithgives given 'Dr/', so the slash form is not Arabic-only. The bareدwith no period is the one shape that still meets the initials question.derek73 commentedon Nov 13, 2026
A related question, now that #1256 gave the parser one table of script properties with a "writes initials" column: should Arabic script have a row saying it writes no initials?
The measurements above already point that way: 0 single-letter-plus-period words in JRC-Names' 17,080 Arabic-script spellings, 0 in
corpus_wikidata_native.jsonl's 212, and none in the King Saud directory, where the letter with a period is always a title (د.,أ.).What the row would change:
د.is refused today because the edge-period fold leaves a bareد, which matches an initial; in a script that writes no initials there is no initial for it to swallow, soد.,أ.andم.could ship as titles withoutড's cost.initials()would give nothing for a name in Arabic script, as it does for Thai and Burmese since Which scripts have case, and which write initials? (ჯეკ MA,ကြည် ကြည် ဌေး,एम. एस. वर्मा) #1256, and the name match would read a single letter as a whole word.nameparser/_policy.pysays the opposite today: its comment lists Arabic among the scripts that "abbreviate" and pinsم. الفارسيas a real initial, with no source recorded.م.before a name is the usual abbreviation ofمهندس, so that example may be a title.What is not yet measured:
ar,faandur, after the copy test that caught Burmese's block being English's.ا.ب. عليreads given 'ا.ب.' today, whereA.B. Alireads given 'A.', middle 'B.'). If the script writes no initials this is an unwritten shape and stays as it is.Hebrew is a separate question: it does write initials with a full stop (
א. ב.before a surname), and also abbreviates with the gershayim.derek73 commentedon Nov 13, 2026
Correction to the evidence above (2026-11-13). The "0 single-letter-plus-period words in JRC-Names' Arabic-script spellings" proves nothing: JRC-Names holds no full stop in any script (0 of its 1,315,051 names, Latin included), so that count could not have come out any other way.
What the corpus can show is a single letter standing as a word of its own, the full stop being absent from the source:
So Arabic-script names in news text carry a lone letter a tenth as often as Latin ones, which still points to "writes no initials", on a count that could have come out otherwise. The 21 have not been read: some may be initials, some the cut-off spellings this issue already noted. The King Saud directory count stands, since that source keeps its full stops.
The Hebrew figure is the opposite signal, and is going into its own issue.
derek73 commentedon Nov 13, 2026
The Hebrew row of the table above is wrong; the Arabic and Latin rows stand. JRC-Names drops apostrophes as well as full stops (no Hebrew name in it holds a geresh, gershayim or apostrophe), so a letter written with a geresh for a foreign sound (
ג',ז',צ') is left standing alone: 394 of the 450 lone letters are those. Without the letters that take a geresh, 89 of 2,684 Hebrew names (3.3%) hold a one-letter word, nearly all in the middle of a three- or four-word name, and nine of them are word-final letter forms, which cannot be initials. So the corpus does not settle Hebrew either way. The same artifact does not touch the Arabic count, Arabic script writing no such mark.