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 recordings in Uzbek and their corresponding transcriptions. The dataset is organized into several folders based on different subsets, such as validated, train, and other datasets.

Dataset Structure

The dataset is structured as follows:

audio/: Contains all audio files.
- uz_other_dataset/: Contains .tar files, each containing multiple .mp3 files.
- uz_train_dataset/: Contains .tar files, each containing multiple .mp3 files.
- uz_validated_dataset/: Contains .tar files, each containing multiple .mp3 files.
- wavs/: Contains individual .wav files.
transcription/: Contains .tsv files with transcriptions.
- other.tsv: Transcriptions for uz_other_dataset.
- train.tsv: Transcriptions for uz_train_dataset.
- validated.tsv: Transcriptions for uz_validated_dataset.
- wavs.tsv: Transcriptions for wavs.

Example Data

Each .tsv file contains the following columns:

file_name: Path to the audio file.
transcription: Corresponding transcription text.

Example:

file_name                                        | transcription
----------------------------------------------- | -----------------------------------------------------------------
audio/uz_validated_dataset/common_voice_uz_30061653.mp3 | Xola, anuv shokoladingizdan obering, dedi chamasi etti yasharlar qiz bidirlab.
audio/uz_validated_dataset/common_voice_uz_28897734.mp3 | Men bu kunni hayotimdagi eng baxtli kun deb hisoblayman.
audio/uz_validated_dataset/common_voice_uz_32191851.mp3 | Qopag‘on itga suyak kerak yoki tayoq.