Skip to content
Fan Zeng (Frank)
Back to resources

Large Bengali ASR Training Corpus

This corpus provides Bengali speech recordings with transcripts and anonymized speaker identifiers.

Visit official source (opens in a new tab)

Research use

Build or adapt a Bengali recognizer and use speaker identifiers when designing evaluation splits.

Limitations

Quality checks can leave residual transcription errors. The release primarily supports supervised speech recognition training.

License and access

Dataset license
CC BY-SA 4.0 (opens in a new tab)
Access status
Direct access

Download the release files from the publisher page and follow the accompanying license and citation terms.