Bayanat Labs Arabic Gazetteer

Arabic gazetteer Arabic place names Arabic geocoding toponym matching GeoNames Arabic

Key numbers

Counted by script from the released files. Each key number is read from meta/stats.json, meta/baseline.json or meta/validation.json in the Hugging Face dataset (the same files are in results/v0.1.0/ in the GitHub repository) and pinned to its SHA-256; MANIFEST.json lists every file with its size and SHA-256.

Downloads and identifiers

Get started

Load a table with the datasets library. To rebuild the gazetteer, install the builder from GitHub at the v0.1.0 tag; the package is not on PyPI yet. The rebuild commands pin the Arab-World Wikidata Slice to revision 6e1295970ec41247d421e4843227f724ae8de32b, the one this release was built from, and the build checks the slice files it reads against the slice's SHA-256 checksums and stops if one differs. GeoNames publishes new files every day, so a rebuild gives somewhat different tables.

# Python
from datasets import load_dataset
places = load_dataset("Bayanat-Labs/arabic-gazetteer", "places", split="train")
names = load_dataset("Bayanat-Labs/arabic-gazetteer", "names", split="train")
test = load_dataset("Bayanat-Labs/arabic-gazetteer", "match_examples", split="test")

# Rebuild
pip install "git+https://github.com/Bayanat-Labs/arabic-gazetteer@v0.1.0"
arabic-gazetteer fetch-geonames geonames/
hf download Bayanat-Labs/arab-world-wikidata --repo-type dataset --revision 6e1295970ec41247d421e4843227f724ae8de32b --local-dir slice/
arabic-gazetteer build --geonames geonames/ --wikidata slice/ --out release/ \
    --geonames-date 2026-10-30 --wikidata-date 2026-10-26 --wikidata-slice-revision 6e1295970ec41247d421e4843227f724ae8de32b

What is released

Attribution: Arabic Gazetteer by Bayanat Labs, CC BY 4.0. Contains data from GeoNames (https://www.geonames.org), licensed under CC BY 4.0, and from Wikidata (CC0 1.0). https://huggingface.co/datasets/Bayanat-Labs/arabic-gazetteer

Intended uses

• Match a location string in Arabic or Latin script, such as a city in a form or a customer address, to one place and its admin units.
• Train and test place-name matchers on the match examples, with same-name places as distractors and a split fixed by admin1 unit.
• Look up the Arabic and Latin-script spellings of a place, and the admin units it belongs to, for search, entity linking or data cleaning.

Out of scope

• Settling names or borders: names, borders and country codes follow the sources as recorded.
• Street-level geocoding: the gazetteer holds places and admin units, not streets or buildings.

Rights at a glance

Training, display, redistribution and synthetic-voice use are separate permissions.

Limitations

• GeoNames and Wikidata are edited by many people and contain errors: wrong or outdated names, misplaced coordinates, duplicate features and wrong links. The records have not been reviewed by people, and the data is provided as is.
• Coverage is uneven: places without an Arabic name in GeoNames or Wikidata are not included, and the share of GeoNames features with an Arabic name is lowest in Sudan and highest in Qatar.
• Lower admin levels are often unresolved in GeoNames; they are left empty, not guessed.
• The match examples are generated by fixed rules from the names in the release; none was labelled by people.
• Flags such as colloquial or historic are the sources' flags, not evidence of dialect or period; population figures come from the sources and may be old.
• Names, borders and country codes follow the sources as recorded; Bayanat Labs takes no position on any boundary. Install the builder from GitHub for now; the package is not on PyPI yet.

Sources and licences

The GeoNames files were downloaded from the GeoNames export server with the Bayanat Labs open-data user agent, one request at a time and more than a second apart; each file's Last-Modified date, size and SHA-256 are recorded in meta/build.json. Wikidata labels, aliases, classes, located-in chains and GeoNames ids come from the Arab-World Wikidata Slice. GeoNames states that its data is provided as is, without warranty of accuracy, timeliness or completeness. Related datasets from Bayanat Labs, linked under Downloads and identifiers: the Arab-World Wikidata Slice, the Wikidata items tied to the 22 states that this gazetteer reads; the Arab World Places Atlas, points of interest, streets and admin areas built from Overture Maps and OpenStreetMap, without this gazetteer's per-place spelling clusters and match examples; and the Arabic Entity Atlas, Arabic names and aliases of Wikidata entities of every kind, linked to their Wikidata ids and not limited to places. Source details

How to cite

Cite as:

Bayanat Labs (2026). Arabic Gazetteer, version 0.1.0 (GeoNames files of 2026-10-30; Wikidata dump of 2026-10-26). Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/arabic-gazetteer

• CITATION.cff

No DOI. Cite the URL in the citation above and the date you read it.

@misc{bayanatlabs2026arabicgazetteer,
  author       = {{Bayanat Labs}},
  title        = {Arabic Gazetteer, version 0.1.0 (GeoNames files of 2026-10-30; Wikidata dump of 2026-10-26)},
  year         = {2026},
  month        = oct,
  howpublished = {\url{https://huggingface.co/datasets/Bayanat-Labs/arabic-gazetteer}}
}

How we handle citations

Changelog

Corrections and removal requests

If something in this release should not be there, or is wrong, email hello@bayanatlabs.com. We handle removal requests within 30 days, and platform requests within 24 hours. How removal requests work

Related product: Bayanat Search