!!! Please also fill out the Google Form to request access to the Wenetspeech dataset.

---

Dataset Card for WenetSpeech

Dataset Description

WenetSpeech is a 10000+ Hours Multi-domain Chinese Corpus for Speech Recognition.

Example Usage

To download the L configuration:

from datasets import load_dataset
ws_train = load_dataset("wenet-e2e/wenetspeech", "L_fixed", split="train", trust_remote_code=True,streaming=True)
# see structure
print(ws_train)

It is possible to download only the development or test data:

ws_dev = load_dataset("wenet-e2e/wenetspeech", "DEV_fixed", split="validation")
ws_test_net = load_dataset("wenet-e2e/wenetspeech", "TEST_NET", split="test")
ws_test_meeting = load_dataset("wenet-e2e/wenetspeech", "TEST_MEETING", split="test")

Licensing Information

The WenetSpeech dataset is available to download for non-commercial purposes under a Creative Commons Attribution 4.0 International License. WenetSpeech doesn't own the copyright of the audios, the copyright remains with the original owners of the video or audio, and the public URL is given for the original video or audio.

Citation Information

Please cite this paper if you find this work useful:

@inproceedings{zhang2022wenetspeech,
  title={Wenetspeech: A 10000+ hours multi-domain mandarin corpus for speech recognition},
  author={Zhang, Binbin and Lv, Hang and Guo, Pengcheng and Shao, Qijie and Yang, Chao and Xie, Lei and Xu, Xin and Bu, Hui and Chen, Xiaoyu and Zeng, Chenchen and others},
  booktitle={ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={6182--6186},
  year={2022},
  organization={IEEE}
}