Repository navigation
Tariq Abdul Aziz: should a two-word surname like Abdul Aziz stay whole after the given name? #694
Description
What happens today. A bound given-name word (rules.md#P5) joins the word after it only at the start of the name. Later in the name it is read as a middle name, in Latin script and native script alike:
Abdul Aziz Tariq → given 'Abdul Aziz', family 'Tariq'
Tariq Abdul Aziz → given 'Tariq', middle 'Abdul', family 'Aziz'
طارق عبد العزيز → given 'طارق', middle 'عبد', family 'العزيز'
Evidence. The native-script Wikidata corpus (#693) holds 150 real Arabic names that align. Every disagreement among them is this one shape, a two-word family name split across middle and family: 7 names, keyed in tools/differential/wikidata_native_agreement.json.
طارق عبد العزيزسمر عبد العزيزنورس عبد الواحدعبد اللطيف عبد الحميدجميل عز الدينهالة شكر اللهعائشة ذو الغنى
Only 4 of the 7 begin with a word in BOUND_GIVEN_NAMES (عبد). The others begin with عز, شكر and ذو. عز الدين and شكر الله are compounds whose SECOND word is the fixed part (الدين "of the religion", الله "God"). So there are two candidate mechanisms:
- (a) a bound word joins forward in any position, not only the first;
- (b) a word list of second elements (
الدين,الله,الغنى) that join BACKWARD to the word before them.
Proposed statement, for (a). A bound given-name word that is not the name's first word joins the word after it, and the pair is placed by position like any one name word. So Tariq Abdul Aziz → family Abdul Aziz.
Boundaries to settle:
Abdul Rahman Abdul Aziz→ givenAbdul Rahman, familyAbdul Aziz.Muhammad Abdul(nothing to join) → familyAbdul, as today.- Family-first order:
Abdul Aziz TariqunderFAMILY_FIRST→ familyAbdul Aziz. - After a family comma:
Abdul Aziz, Tariqalready reads familyAbdul Aziz; check it is unchanged.
Accepted consequences, if (a):
- A three-word name whose middle word is
Abdulnow reads it as part of the family. IfAbdul Azizwere a second given name in a patronymic chain, that reading would be wrong. - Nasab chains (
X bin Abdul Aziz) need checking against P2'sbinchain.
Open questions:
- Does (b) belong in the default vocabulary, or only in a future Arabic pack (see the v3 language-packs design notes)?
- Does the join in non-leading position need the same "a word to spare" reserve P5 uses?
Exclusions. The Persian (ییلاقی اشرفی) and Bulgarian (Цветков Дерменджиев) double surnames in the same record are a different shape: two family words with no bound word. They belong to the double-surname question, not this one.
Activity
derek73 commentedon Nov 4, 2026
Measured: would a particle reading of
abdulsettle the Latin-script half?A particle in the middle of a name chains forward into the family (rules/P2), so giving
abdula particle reading readsTariq Abdul Azizas familyAbdul Aziz.Hassan Abu Talibalready reads that way, sinceabuis both a bound given-name word and a teknonymic particle (#269). That is option (a) for Latin script, with no new mechanism. The build requires a particle that is also a bound given-name word to have agivenreading too, so a name opening withAbdulwould start reportingPARTICLE_OR_GIVEN, asAbudoes.Where filers put a non-leading
Abdulwith a word behind it (2026-11-04; ORCID's 2025 public file as fetched 2026-10-27, and the UK PSC register as fetched 2026-10-22):Abdulin given/middle, next word the surnameThe sources disagree:
Ahmad Abdul Rahmanwrites the father's name withoutbin.Mohammed Abdul Khan, withAbdula middle name, which the default reads right today and a particle reading would make familyAbdul Khan.abusplits the same way (ORCID 1,209 family against 144 otherwise; PSC 75 against 95), so its particle reading already pays this cost.What it does not reach:
عز الدين,شكر الله,ذو الغنى). Their fixed part is the second word, which is option (b).عبدwould needعبدto read as a particle too.Proposal: not a default. The convention is Malay and Arab, and British South Asian filers write the other way. A particle reading of
abdul(andعبد) writtendefault = false, which a Malay or Arabic language choice turns on, would follow howesandcscarry their family-name conventions. Next step: measure that choice per country withtools/differential/config_diff.pyagainst ORCID and PSC. Option (b) stays open.Recompute: over each record's boxes in order (ORCID given then family; PSC forename, middle_name, surname), take the first
Abdul/Abu(case-folded) that is not the first word and has a word after it, and key it by the boxes holding it and the next word.