metadata

dataset_info:
  features:
    - name: id
      dtype: string
    - name: audio
      dtype: audio
    - name: sentence
      dtype: string
  splits:
    - name: train
      num_bytes: 12400028580.739
      num_examples: 100767
    - name: validation
      num_bytes: 519447209.472
      num_examples: 3783
    - name: test
      num_bytes: 596361411.064
      num_examples: 3837
  download_size: 11656859636
  dataset_size: 13515837201.275
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
license: cc-by-4.0
task_categories:
  - automatic-speech-recognition
language:
  - uz
tags:
  - ISSAI
  - USC
  - STT
pretty_name: ISSAI_USC
size_categories:
  - 100K<n<1M

Uzbek Speech Corpus

Dataset Description
Dataset Structure
Dataset Creation
Considerations for Using the Data
Additional Information

Dataset Description

Homepage: https://issai.nu.edu.kz/uzbek-asr/
Repository: https://github.com/IS2AI/Uzbek_ASR
Paper: https://arxiv.org/abs/2107.14419
Point of Contact: [email protected]
Size of downloaded dataset files: 11.66 GB
Size of the generated dataset: 1.86 GB
Total amount of disk used: 13.52 GB

Dataset Summary

The Uzbek speech corpus (USC) has been developed in collaboration between ISSAI and the Image and Speech Processing Laboratory in the Department of Computer Systems of the Tashkent University of Information Technologies. The USC comprises 958 different speakers with a total of 105 hours of transcribed audio recordings. To ensure high quality, the USC has been manually checked by native speakers. The USC is primarily designed for automatic speech recognition (ASR), however, it can also be used to aid other speech-related tasks, such as speech synthesis and speech translation. To the best of our knowledge, the USC is the first open-source Uzbek speech corpus available for both academic and commercial use under the Creative Commons Attribution 4.0 International License. We expect that the USC will be a valuable resource for the general speech research community and become the baseline dataset for Uzbek ASR research.

Please refer to paper and GitHub repository for further details.

Disclaimer: Abror Shopulatov, who was not involved in this research, converted the original dataset and wrote the contents of this model card based on the original paper and blogpost. This is HuggingFace version of the dataset that is created for mainly easy to access usage and can be transfered to original creators upon request.

Dataset Structure

Data Instances

Size of downloaded dataset files: 11.66 GB
Size of the generated dataset: 1.86 GB
Total amount of disk used: 13.52 GB

Data Fields

The data fields are the same among all splits.

id: a string feature to indicate id of the given audio and sentence pair.
audio: an audio feature to store audio.
sentence: a string feature to indicate normalized version of the text of corresponding audio.

Data Splits

Category	train	validation	test	total
Duration (hours)	96.4	4.0	4.5	104.9
Utterances	100,767	3,783	3,837	108,387
Words	569.0k	22.5k	27.1k	618.6k
Unique Words	59.5k	8.4k	10.5k	63.1k
Speakers	879	41	38	958

Citation Information

@misc{musaev2021uscopensourceuzbekspeech,
      title={USC: An Open-Source Uzbek Speech Corpus and Initial Speech Recognition Experiments}, 
      author={Muhammadjon Musaev and Saida Mussakhojayeva and Ilyos Khujayorov and Yerbolat Khassanov and Mannon Ochilov and Huseyin Atakan Varol},
      year={2021},
      eprint={2107.14419},
      archivePrefix={arXiv},
      primaryClass={eess.AS},
      url={https://arxiv.org/abs/2107.14419}, 
}

Contact

For any questions or issues related to the dataset or code, please contact [email protected].

Datasets:

murodbek
/

uzbek-speech-corpus