Quran Muaalem

English

العربية

English

العربية

Appearance

Quran MuaalemCorrect Holy Quran recitations by analyzing phonemes, diacritization, articulatory attributes, and Tajweed rules using open‑source AI

Model + tools + transcript pipeline in one repo

Getting Started
Quran Transcript

Multi‑level CTC

The model decodes phonemes and Tajweed‑related attributes using a multi‑level CTC head.

Reference‑aware analysis

Inference compares predicted phonemes with a reference phonetic script from quran-transcript.

Gradio UI

A ready‑to‑run UI is wired in src/quran_muaalem/gradio_app.py.

What this repo contains ​

  • Quran Muaalem inference and UI code under src/quran_muaalem/.
  • A full Quran transcript toolkit under quran-transcript/ (packaged as quran-transcript).
  • Deployment and experiment helpers under deploy/, tests/, and assets/.

For researchers: where to start ​

If your goal is evaluation or reproducibility:

  1. Read Architecture to understand multi‑level CTC.
  2. Review Pipeline Steps for label generation.
  3. Use Evaluation and Metrics to standardize reporting.
  4. Check Outputs for the exact output schema.

Practical notes ​

  • The model expects 16 kHz audio.
  • Output quality depends on the reference phonetic script produced by quran_transcript.quran_phonetizer.
  • Recommended metrics and reporting templates live under Evaluation and Metrics.

Use the navigation to dive into each component.