Skip to content

Dataset · AISHELL

Open Chinese speech corpus for ASR

AISHELL/AISHELL-1

Studio speech recordings of 400 Mandarin speakers with manual transcription accuracy above 95% for speech recognition tasks.

What's inside

The dataset contains Mandarin Chinese speech audio recordings captured in a quiet environment using high-quality microphones and resampled to 16 kHz. It features 400 speakers from various dialect regions of China, with manual transcription accuracy exceeding 95%. Published by Beijing Shell Shell Technology and open for academic use.

What you can do with it

  • Train and evaluate Mandarin automatic speech recognition (ASR) models
  • Test speech model robustness against various regional Chinese accents
Tasks
Automatic speech recognition
Languages
Chinese
Downloads this month
9.4K
Likes
20

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("AISHELL/AISHELL-1")