AliMeeting
118.75-hour Chinese multi-channel meeting speech dataset from Alibaba supporting speaker diarization and multi-speaker ASR
Duration
119 hours
Languages
1
Sample Rate
16 kHz
Published
2022-01
Description
1Recordings from real meeting scenarios
2Total 118.75 hours: 104.75h training, 4h validation, 10h test
3Covers various meeting room environments, different numbers of participants, and varying speaker overlap ratios
4Recorded using 8-channel microphone arrays and headset microphones
5Designed for the ICASSP 2022 M2MeT Challenge
6Supports speaker diarization and multi-speaker ASR tasks
Language Details
| Language | Duration |
|---|---|
| Mandarin Chinese | 119 hours |
Publisher