Bayanat Labs Arab World Places Atlas

Arabic points of interest Arabic gazetteer Arabic geocoding Arabic street names Arabic transliteration Overture Maps Arabic Middle East POI dataset

Key numbers

Counted by script from the released files. Each key number is read from stats.json in the Hugging Face dataset and pinned to its SHA-256; a checksum file in the dataset lists the SHA-256 of every file.

Downloads and identifiers

Get started

Load any of the six configurations with the datasets library, or a single country's places file. You can also query the Parquet files directly with DuckDB.

from datasets import load_dataset
repo = "Bayanat-Labs/arab-world-places-atlas"
places = load_dataset(repo, "places", split="train")
saudi = load_dataset(repo, data_files="data/places/places-SA.parquet", split="train")
pairs = load_dataset(repo, "geo_name_pairs")  # train / validation / test

What is released

Attribution: Arab World Places Atlas by Bayanat Labs (2026), built from Overture Maps Foundation data (release of 21 October 2026); © OpenStreetMap contributors. https://huggingface.co/datasets/Bayanat-Labs/arab-world-places-atlas

Intended uses

• Match a typed Arabic or Latin-script place name to candidate places, and rank them by country, region, county and H3 cell.
• Build street, district and city vocabularies for Arabic address parsers, gazetteers and form auto-complete.
• Train or evaluate Arabic–Latin transliteration and name translation on real map names with fixed splits.
• Compare category density across countries, regions and H3 cells.

Out of scope

• Treating it as a complete or verified business register: entries can be closed, duplicated, miscategorized or misplaced.
• Contacting businesses or individuals: phone numbers, e-mail addresses and social-media links are not included.
• Drawing conclusions about borders: country assignment follows Overture's published divisions.

Rights at a glance

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

Limitations

• Coverage is urban-heavy and comes mostly from one contributing platform; Comoros, Djibouti, Mauritania, Somalia, Sudan and Yemen are thin.
• Only 396,783 places carry an Arabic-script name; many Gulf listings are English-only at source.
• Most places have a low Overture confidence score; filter on the confidence field for high-precision uses.
• The Arabic and Latin halves of a bilingual business name are not always translations or transliterations of each other.
• Overture publishes Palestine as two areas (XW and XG), shown here as PS, and includes Western Sahara within Morocco; Bayanat Labs takes no position on any boundary.

Sources and licences

Admin assignment, name splitting, Arabic search keys, street grouping, name pairs and splits are by Bayanat Labs. The underlying place, street and boundary data belong to Overture Maps Foundation, its providers and OpenStreetMap contributors. Phone numbers, e-mail addresses and social-media links were removed. Source details

How to cite

Cite as:

Bayanat Labs (2026). Arab World Places Atlas, version 1.0.0. Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/arab-world-places-atlas

• CITATION.cff

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

@misc{bayanatlabs2026arabworldplacesatlas,
  author       = {{Bayanat Labs}},
  title        = {Arab World Places Atlas, version 1.0.0},
  year         = {2026},
  month        = oct,
  howpublished = {\url{https://huggingface.co/datasets/Bayanat-Labs/arab-world-places-atlas}}
}

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