AISHELL-3

Dataset Description

AISHELL-3 is a large-scale and high-fidelity multi-speaker Mandarin speech corpus published by Beijing Shell Shell Technology Co., Ltd.

The dataset is designed for training multi-speaker Text-to-Speech (TTS) systems. It contains roughly 85 hours of emotion-neutral recordings spoken by 218 native Mandarin Chinese speakers, with a total of 88,035 utterances.

Auxiliary speaker attributes, including gender, age group, and native accents, are explicitly marked and provided in the corpus.

Transcripts are provided at both:

• Chinese character level
• Pinyin level

The word and tone transcription accuracy rate is above 98%, based on professional speech annotation and strict quality inspection for tone and prosody.

Dataset Source

• OpenSLR: http://openslr.org/93/
• External URL: http://www.aishelltech.com/aishell_3

Dataset Download

The dataset is available from OpenSLR.

File	Size	Description
data_aishell3.tgz	19 GB	Speech data and transcripts

Dataset Information

Item	Description
Identifier	SLR93
Category	Speech
Language	Mandarin Chinese
Publisher	Beijing Shell Shell Technology Co., Ltd.
Duration	Roughly 85 hours
Number of speakers	218
Number of utterances	88,035
Speech style	Emotion-neutral
Transcript type	Chinese character-level and pinyin-level transcripts
License	Apache License v2.0

Intended Uses

AISHELL-3 can be used for research and development of:

• Multi-speaker Text-to-Speech systems
• Mandarin speech synthesis
• Speaker-conditioned speech generation
• Speech corpus analysis for Mandarin TTS

License

This dataset is released under the Apache License v2.0.

Citation

If you use this dataset, please cite the following reference:

@inproceedings{AISHELL-3_2020,
  title={AISHELL-3: A Multi-speaker Mandarin TTS Corpus and the Baselines},
  author={Yao Shi, Hui Bu, Xin Xu, Shaoji Zhang, Ming Li},
  year={2015},
  url={https://arxiv.org/abs/2010.11567}
}