|
|
--- |
|
|
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. |
