AISHELL-Stammertalk

Name: AISHELL-Stammertalk
Creator: Beijing University of Posts and Telecommunications, Northwestern Polytechnical University, Beijing AISHELL Technology Co., Ltd., China Rehabilitation Research Center, Nankai University, Shanghai Jiao Tong University, University of Science and Technology of China, Duke Kunshan University
Published: 2024-06

48.8-hour Mandarin stuttered speech dataset for ASR and stuttering event detection

Duration

48.8 hours

Languages

Sample Rate

None

Published

2024-06

Description

1Also known as AS-70, this is a professional Mandarin stuttered speech dataset

2Collected from 70 native Mandarin adults who stutter, including 46 males and 24 females

3The total duration is 48.8 hours, with each participant contributing about a one-hour recording session

4Recordings include free conversation and voice command reading

5Free conversations were conducted as online interviews through platforms such as Zoom and Tencent Meeting, capturing spontaneous speech across diverse topics

6Voice command reading covers in-car navigation and smart-home interaction scenarios, with 600 unique commands in total

7Annotates five stuttering event types: word/phrase repetition, block, prolongation, sound repetition, and interjection

8Designed for stuttered speech detection, recognition, and related speech technology research

Language Details

Language	Duration
Mandarin Chinese	48.8 hours

Publisher

Beijing University of Posts and TelecommunicationsNorthwestern Polytechnical UniversityBeijing AISHELL Technology Co.Ltd.China Rehabilitation Research CenterNankai UniversityShanghai Jiao Tong UniversityUniversity of Science and Technology of ChinaDuke Kunshan University

License & Commercial Use

License

None

Commercial Use

Unknown

Resources

Paperhttps://arxiv.org/abs/2406.07256 Datasethttps://opendata.aishelltech.com/stammertalk Official Pagehttps://www.aishelltech.com/AISHELL_6A Sample / Demohttps://stammertalk.github.io/interspeech2024-page/