Downloading publicly available datasets

We provide scripts to download and process the following publicly available datasets:

• An4 - Alphanumeric database
• Librispeech - reading english books
• TED-LIUM 3 (ted3) - TED talks
• Voxforge
• common voice (old version)

Simply run the respective scripts in sonosco > datasets > download_datasets with the output_path flag and it will download and process the dataset. Further, it will create a manifest file for the dataset.

For example

python an4.py --target-dir temp/data/an4