Speech datasets obtained from the Common Voice project:

• Dutch

  • Common Voice (10.0-2022-07-04) (2.7GB)

• Indonesian

  • Common Voice (10.0-2022-07-04) (1.3GB)

  • Coqui STT (10.0-2022-07-04) (1.4GB), conversion

• Japanese

  • Common Voice (10.0-2022-07-04) (1.1GB)

  • Coqui STT (10.0-2022-07-04) (1.8GB), conversion

• Norwegian Nynorsk

  • Common Voice (10.0-2022-07-04) (18.5MB)

  • Coqui STT (10.0-2022-07-04) (61.5MB), conversion

Relevant datasets to use from the archives:

• Common Voice

  • train.tsv - the training set

  • dev.tsv - the validation set

  • test.tsv - the test set

• Coqui STT

  • train.csv - the training set

  • dev.csv - the validation set

  • test.csv - the test set

Conversion into other formats can be achieved with the wai.annotations library.

License

CC-0