---
language:
- ar
size_categories:
- 100K<n<1M
---

TLOG is a dataset consisting of audio recitations of Quranic Ayahs and their corresponding Quranic texts.

### Features
Each data point of this dataset consists of the following features:
- `audio`: audio recitation of a specific Ayah in the Quran
  - for a clean data point:
    - `array`: the audio signal in array form
    - `sample_rate`: the sample rate of the audio signal
    - `path`: the file name, which, for a clean data point, should correspond to the verse that’s being recited:
      - The format is: `surahNum_ayahNum_id.wav`
      - For example, `2_1_2321423.wav` corresponds to the text in Surah 2, Verse 1.
  - for an unclean data point:
    - `bytes`: the raw audio bytes, un-decoded, as it could be corrupt
    - `path`: the file name
- `label`: Arabic text of the Ayah recited, in Uthmani script with Tashkeel. If data point is unclean, label is 0.
- `is_clean`: indicates whether the specific data point is clean or unclean.

### Splits
- `clean`: audio files that follow a valid filename format and is not corrupt
- `unclean`: audio files that do not follow a valid filename format and/or is corrupt

### Note
The labels were inferred from the Surah/Ayah number of the audio filename and are not validated further.