Bayanat Labs GCC Format Validators and Arabic ITN/TN Test Pairs
A speech recogniser may write an amount out in words where the transcript should show digits, and a text-to-speech front end has to read digits, currency symbols and dates aloud. This release gives speech, document and fintech teams test pairs for both directions, inverse text normalisation (ITN) and text normalisation (TN), in Modern Standard Arabic, Saudi and Gulf Arabic and Egyptian Arabic, with Western and Arabic-Indic digits: cardinals, ordinals, money, Hijri and Gregorian dates, times, phone numbers, identity numbers, IBANs, percentages and measures. With them come reference tables and validators for the formats the pairs contain: Saudi, Emirati and Egyptian identity numbers, Kuwaiti and Bahraini identity numbers (structure only), IBAN structures and phone numbering plans of Arab League states, Saudi plate letters, an Umm al-Qura Hijri to Gregorian table cross-checked against ICU, and Arabic number and currency words. All identifiers, IBANs and phone numbers are generated values that pass the format checks; they are not real personal data.
| Licence | CC BY 4.0 (test pairs and tables); CC0 1.0 (format facts); Apache-2.0 (phone plans and code) |
|---|---|
| Version | v0.1.0 · Test pairs, tables and validators v0.1.0, tagged on GitHub. The package is not on PyPI yet. |
| Published | |
| Updated | |
| Who it is for | Speech-recognition, text-to-speech, document and fintech teams that convert Arabic numbers, amounts, dates and identifiers between spoken and written form, or check the format of Gulf identifiers, IBANs and phone numbers. |
| Level | Intermediate |
Key numbers
Counted by script from the released files. Each key number is read from qa/stats.json in the Hugging Face dataset and pinned to its SHA-256.
Downloads and identifiers
| Where | What |
|---|---|
| Hugging Face | Dataset on Hugging Face |
| GitHub | Validators and scorer on GitHub |
Get started
Load the test pairs with the datasets library, or install the validators and scorer from GitHub at the v0.1.0 tag (not on PyPI yet).
# Python
from datasets import load_dataset
pairs = load_dataset("Bayanat-Labs/gcc-format-validators", "itn_tn_pairs", split="test")
# Validators and scorer
pip install "git+https://github.com/Bayanat-Labs/gcc-format-validators@v0.1.0"
gcc-formats score --gold test.jsonl --pred predictions.jsonl --direction itnWhat is released
| Part | Licence | In this release |
|---|---|---|
| ITN/TN test pairs (dev and test splits) | CC BY 4.0 | Yes |
| Compiled format tables, calendar table and number words | CC BY 4.0 | Yes |
| Format facts (lengths, prefixes, field positions, check-digit rules, letter maps) | CC0 1.0 | Yes |
| Phone numbering plans (from libphonenumber metadata) | Apache-2.0 | Yes |
| Validators and scorer (Python package) | Apache-2.0 | Yes |
Attribution: GCC Format Validators and Arabic ITN/TN Test Pairs by Bayanat Labs, CC BY 4.0. https://huggingface.co/datasets/Bayanat-Labs/gcc-format-validators
Intended uses
- Test an ITN or TN system per class, Arabic variety and digit script, with exact, variant and normalised accuracy from the scorer.
- Check the format of Saudi, Emirati and Egyptian identity numbers, IBANs, phone numbers and Saudi plates.
- Convert between Umm al-Qura Hijri and Gregorian dates, and write Arabic numbers and amounts in words.
Out of scope
- Checking whether an identifier, IBAN or phone number belongs to a real person or account: the validators check the format only.
- Training on the test split: it is meant for evaluation.
Rights at a glance
Training, display, redistribution and synthetic-voice use are separate permissions.
| Use | Status | Note |
|---|---|---|
| Training | Allowed | CC BY 4.0 for the pairs and compiled tables; format facts under CC0 1.0; upstream notices kept for derived fields. |
| Display | Allowed | CC BY 4.0, with attribution. All identifiers are generated values, not personal data. |
| Redistribution | Allowed | CC BY 4.0 with attribution; phone plans under Apache-2.0; MIT and Unicode License V3 notices kept for derived fields. |
| Synthetic voice | Not applicable | Text and tables only. |
Limitations
- The test pairs have not yet been reviewed by native speakers of each variety; every pair is labelled with its review status.
- Kuwaiti and Bahraini identity numbers are covered for structure only.
- The Umm al-Qura table and ICU's islamic-umalqura calendar agree on every day in a central span of years and differ on many days outside it; every difference is listed in the dataset, with the counts.
- No system baseline is published in this version.
Sources and licences
| Source | Licence | Redistributed here |
|---|---|---|
| libphonenumber metadata | Apache-2.0 | Yes |
| schwifty IBAN registry data | MIT | Yes |
| hijridate (Umm al-Qura month starts) | MIT | Yes |
| Unicode CLDR 48.2 (currency names) | Unicode License V3 | Yes |
Test pairs, templates and compiled tables are by Bayanat Labs. Fields derived from the sources above keep their upstream notices (NOTICE in the repository and LICENSE-sources.md in the dataset); ISO currency codes and minor units are facts published by SIX. Source details
How to cite
Cite as:
Bayanat Labs (2026). GCC Format Validators and Arabic ITN/TN Test Pairs, version 0.1.0. Bayanat Labs. https://huggingface.co/datasets/Bayanat-Labs/gcc-format-validators
No DOI. Cite the URL in the citation above and the date you read it.
@misc{bayanatlabs2026gccformatvalidators,
author = {{Bayanat Labs}},
title = {GCC Format Validators and Arabic ITN/TN Test Pairs, version 0.1.0},
year = {2026},
month = oct,
howpublished = {\url{https://huggingface.co/datasets/Bayanat-Labs/gcc-format-validators}}
}
Changelog
| Date | Version | Change |
|---|---|---|
| v0.1.0 | Test pairs, format tables, calendar table, validators and scorer on GitHub and Hugging Face. |
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 Voice