Data Sources

This project references external tools and datasets in README.md. The main sources used in the pipeline are:

• Recitation collection: https://github.com/obadx/prepare-quran-dataset
• Segmentation by pauses: https://github.com/obadx/recitations-segmenter
• Quran‑tuned Whisper model: https://huggingface.co/tarteel-ai/whisper-base-ar-quran
• Correction + script conversion: https://github.com/obadx/quran-transcript

How these sources fit together

1. Raw audio is collected and curated.
2. Segments are created using pause‑based splitting.
3. Automatic transcription provides initial Imlaey text.
4. Tasmeea correction improves transcription fidelity.
5. Script conversion yields Uthmani text.
6. Phonetizer generates phoneme + sifat labels.

Recommended metadata to track

For reproducibility, record these fields per segment:

• Reciter id / source
• Recitation style (murattal/mujawad/hadr)
• Audio format (sample rate, bitrate)
• Segment boundaries (start/end timestamps)
• Reference sura/ayah and word span
• Moshaf attributes used

Licensing note

Each external dataset or model has its own license. Record and cite the original source licenses in your dataset documentation.