akhanafer/arabic-to-arabizi

Arabic to Arabizi (Levantine)

A small parallel dataset of short, everyday Levantine (Lebanese) Arabic sentences written in Arabic script, each paired with an Arabizi (Latin-script) transliteration.

Dataset structure

Each row has two fields:

• —arabic: the sentence in Arabic script
• —arabizi: the same sentence in Arabizi

Example:

Transliteration conventions

• —3 = ع
• —2 = ء / أ (hamza)
• —sh = ش, kh = خ, gh = غ
• —h is used for both ح and ه (7 is not used)
• —q = ق
• —Short vowels are mostly omitted; long vowels are written (a, y/i, ou/o)

How the data was created

The Arabic sentences and their Arabizi transliterations were generated with OpenAI's GPT-4, then reviewed by the author.

Intended uses

Training or evaluating Arabic ↔ Arabizi transliteration, input methods and keyboards, and dialect-aware NLP.

Limitations

• —Small size (433 pairs); short, conversational sentences only.
• —Lebanese/Levantine dialect only; not representative of other dialects or Modern Standard Arabic.
• —Reflects one Arabizi spelling style. Real-world Arabizi varies a lot between writers (e.g. 7 for ح, 5 for خ, more written vowels).

License

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. You may use, share and adapt it, including commercially, as long as you give appropriate credit.

Citation

If you use this dataset, please credit:

akhanafer, "Arabic to Arabizi (Levantine)", Hugging Face, 2026.
https://huggingface.co/datasets/akhanafer/arabic-to-arabizi