VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation
This corpus provides a valuable resource for researchers in speech representation learning and semi-supervised learning, particularly for multilingual and low-resource settings, by offering the largest open dataset of its kind.
This paper introduces VoxPopuli, a large-scale multilingual speech corpus comprising 100K hours of unlabelled speech data across 23 languages, along with 1.8K hours of transcribed speeches and 5.1K hours of aligned oral interpretations. The authors provide speech recognition baselines and demonstrate the utility of the unlabelled data for semi-supervised learning in out-of-domain scenarios.
We introduce VoxPopuli, a large-scale multilingual corpus providing 100K hours of unlabelled speech data in 23 languages. It is the largest open data to date for unsupervised representation learning as well as semi-supervised learning. VoxPopuli also contains 1.8K hours of transcribed speeches in 16 languages and their aligned oral interpretations into 5 other languages totaling 5.1K hours. We provide speech recognition baselines and validate the versatility of VoxPopuli unlabelled data in semi-supervised learning under challenging out-of-domain settings. We will release the corpus at https://github.com/facebookresearch/voxpopuli under an open license.