NahwAI/arabic-tashkeel-speech

Nahw Arabic Tashkeel Speech Dataset

An open-source collection of 1,093 fully diacritized Arabic speech recordings, crowd-sourced from native speakers via Nahw.ai.

Dataset summary

Stat	Value
Total recordings	1,093
Speakers	10
Language	Arabic (ar)
Sampling rate	16 kHz
License	CC-BY-4.0

Features

• audio: The speech recording, resampled to 16 kHz.
• transcription: The fully diacritized Arabic sentence that was read aloud.
• sentence: The same sentence without diacritics (tashkeel removed).
• speaker_id: An anonymized speaker identifier.

Usage

from datasets import load_dataset

ds = load_dataset("NahwAI/arabic-tashkeel-speech")
print(ds["train"][0])

Data collection

Native Arabic speakers recorded sentences through the Nahw.ai platform. Each recording was reviewed and approved by a human annotator before inclusion in this dataset.

License

This dataset is released under the Creative Commons Attribution 4.0 International License.

Citation

@dataset{nahw_arabic_speech_2026,
  title={Nahw Arabic Tashkeel Speech Dataset},
  author={Nahw.ai},
  year={2026},
  url={https://huggingface.co/datasets/NahwAI/arabic-tashkeel-speech},
  license={CC-BY-4.0}
}