Skip to content

Where should an Ethiopian father's name go when there is no family name? (Abiy Ahmed Ali gives middle Ahmed; Abiy Ahmed gives family Ahmed) #1203

Open

Description

@derek73

Ethiopian and Eritrean names have no family name. A person is named by their own given name, then their father's given name, then often their grandfather's (Abiy Ahmed Ali). Abroad, the father's name is filed as the surname. The parse reads by position, so the father's name lands in middle or family depending on whether the grandfather is written (measured on master ccaf6ad):

Input setting given middle family
Abiy Ahmed Ali default Abiy Ahmed Ali
Abiy Ahmed default Abiy '' Ahmed
Abiy default Abiy '' ''
Ali, Abiy Ahmed default Abiy Ahmed Ali
አብይ አህመድ አሊ default አብይ አህመድ አሊ
Abiy Ahmed Ali middle_as_family Abiy '' Ahmed Ali
Abiy Ahmed Ali two_given_names Abiy Ahmed '' Ali

patronymic is '' in every row. The lone word reads as the given name with a given-or-family ambiguity (O5, the lone name word's convention), which is the right reading here.

Measured (2026-11-08):

  • House of Peoples' Representatives (OpenSanctions et_hopr, 2,291 distinct names): 2,039 of three words, 211 of two, 41 of four or more. The default reads 2,037 of the three-word names as own name / father (middle) / grandfather (family). The list has no fields, so the expected reading comes from the convention.
  • ORCID's 20,000 Ethiopian records, where holders filled in their own given and family boxes:
    • 11,913 file one given word and one family word, so the father's name is the family.
    • Three-word filings split 5,477 with the father's name in the given box (Lemessa Benti / Merga shape) against 1,384 with it in the family box beside the grandfather's.
    • config_diff.py score default 'default@middle_as_family=1' --group ET --group ER: ET 91.99% by default against 70.46% (18,794 records); ER 92.24% against 72.41% (232 records).
  • Wikidata: the family name (P734) is filled for 56 of 445 Ethiopians, and in 45 it is the last word of the English label, mostly two-word names, so the father's name.

So 80% of the people who split their own three-word name put the father's name with their own, as the default reads it. But the same person filed with two words gets the father's name as family and filed with three words gets the grandfather's, so family doesn't match across their records.

The options

Option What it gives Cost
A. Keep it positional Agrees with 80% of self-filed three-word records and with every two-word one family holds the father in one name and the grandfather in another
B. Father's name as family, as foreign filing does O3 (compound family names written unmarked) already gives middle_as_family: family Ahmed Ali −21.5 points on ORCID ET; the grandfather has no field of his own
C. A patronymic view: an opt-in rule (an am/ti/om pack) fills it with the second name unit, and no field moves The father's name is in one place whatever the word count Breaks R9's stated basis: "The view reads only a word or a rule that says patronymic, never a position", and its boundary that a patronymic with no particle and no rule to read it is not read. A pack's choice would have to declare the convention, as RU's does for O1's endings. Without the compound join (#1201), መንግስቱ ኃይለ ማርያም would give the view ኃይለ alone
D. B and C together Both costs

The closest cases on record:

Questions

  • Is a position a strong enough signal for a pack to read, where every patronymic rule so far reads a word?
  • If C: should the grandfather's name get a view too, or stay in family?
  • Does a caller matching records need anything beyond C? Two records of one person (Abiy Ahmed, Abiy Ahmed Ali) would agree on given and patronymic but not on family.

Metadata

Metadata

Assignees

Labels

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions