Hugging Face's logo

Datasets:
Beehzod
/
STT_uz

Tasks:
Automatic Speech Recognition
Modalities:
Audio
Text
Formats:
webdataset
Languages:
Uzbek
Size:
10K - 100K
Tags:
audio
Libraries:
Datasets
WebDataset
License:
Dataset card Viewer Files Community
3
STT_uz / README.md
Beehzod's picture
Beehzod
Update README.md
767c5cb verified
preview code
|
raw
history blame
2.72 kB
metadata
license: apache-2.0
task_categories:
  - automatic-speech-recognition
language:
  - uz
pretty_name: Uzbek_language_stt_dataset
size_categories:
  - 100K<n<1M

Uzbek Speech-to-Text (STT) Dataset

This dataset contains audio files and their corresponding transcriptions in Uzbek language, organized into different folders based on data splits. The dataset is designed to support speech recognition tasks in Uzbek.

Dataset Structure

The dataset is organized into the following main directories within the audio/ folder:

1. wavs/

This folder contains individual .wav audio files. These files are typically short audio clips. The corresponding transcriptions are provided in the metadata.csv file.

  • Number of files: 1,171
  • File format: .wav
  • Example file path: audio/wavs/file_299-0dae1d85fd1e109ddca5d1d394ff18cf.wav

2. uz_other_dataset/

This folder contains audio data stored in .tar archives. Each .tar file contains multiple audio clips in .mp3 format. The transcriptions for these files are also included in the metadata.csv.

  • Number of files: 4 .tar archives
  • File format: .mp3 inside .tar
  • Example file path: audio/uz_other_dataset/uz_other_0.tar

3. uz_validated_dataset/

Similar to uz_other_dataset, this folder contains validated audio data stored in .tar archives. These files have passed quality checks and are validated transcriptions.

  • Number of files: 3 .tar archives
  • File format: .mp3 inside .tar
  • Example file path: audio/uz_validated_dataset/uz_validated_0.tar

4. uz_train_dataset/

This folder contains training data stored in .tar archives. The transcriptions for these files are aimed at building robust speech recognition models.

  • Number of files: 2 .tar archives
  • File format: .mp3 inside .tar
  • Example file path: audio/uz_train_dataset/uz_train_0.tar

Metadata

The dataset includes a metadata.csv file that provides the following columns:

  • file_name: The path to the audio file relative to the audio/ folder.
  • transcription: The corresponding transcription for the audio file.

Example Row 

file_name,transcription
audio/wavs/file_299-0dae1d85fd1e109ddca5d1d394ff18cf.wav,qabul davomida 16 nafar fuqaroning 17 ta turli masalalardagi murojaatlari ko'rib chiqildi
audio/uz_other_dataset/common_voice_uz_28889488.mp3,Ammo yig‘lashga arziydigan xato bor.
audio/uz_validated_dataset/common_voice_uz_30538165.mp3,Toshkent va Sirdaryoda yer maydonini sotmoqchi bo‘lgan firibgarlar ushlandi
audio/uz_train_dataset/common_voice_uz_40010123.mp3,Toshkent va Sirdaryoda yer maydonini sotmoqchi bo‘lgan firibgarlar ushlandi