Skip to content
Fan Zeng (Frank)
Back to resources

Multilingual LibriSpeech (MLS)

Multilingual LibriSpeech builds parallel dataset structure across eight languages from LibriVox recordings.

Visit official source (opens in a new tab)

Research use

Use language-specific splits to compare transfer learning or the effect of unequal training volume.

Limitations

English accounts for most of the collection. Measure each language separately to make the imbalance visible.

License and access

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

Use the publisher’s resource page and check the terms for the files and version you choose.

Additional sources