Add VoxLingua107 dataset

[osanseviero]

Adding a Dataset

• Name: VoxLingua107
• Description: VoxLingua107 is a speech dataset for training spoken language identification models. The dataset consists of short speech segments automatically extracted from YouTube videos and labeled according the language of the video title and description, with some post-processing steps to filter out false positives.
• Paper: https://arxiv.org/abs/2011.12998
• Data: http://bark.phon.ioc.ee/voxlingua107/
• Motivation: Nice audio classification dataset

cc @anton-l

Instructions to add a new dataset can be found here.

[polinaeterna]

self-assign