VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation

Changhan Wang (1) , Morgane Rivière (1) , Ann Lee (1) , Anne Wu (1) , Chaitanya Talnikar (1) , Daniel Haziza (1) , Mary Williamson (1) , Juan Pino (2) , Emmanuel Dupoux (3, 4, 1, 5, 6, 7)

1 FAIR - Facebook AI Research [Paris]
2 Chercheur indépendant
3 EHESS - École des hautes études en sciences sociales
4 ENS-PSL - École normale supérieure - Paris
5 DEC - Département d'Etudes Cognitives - ENS Paris
6 CoML - Apprentissage machine et développement cognitif
7 LSCP - Laboratoire de sciences cognitives et psycholinguistique

We introduce VoxPopuli, a large-scale multilingual corpus providing 400K hours of unlabeled speech data in 23 languages. It is the largest open data to date for unsupervised representation learning as well as semisupervised learning. VoxPopuli also contains 1.8K hours of transcribed speeches in 15 languages and their aligned oral interpretations into 15 target languages totaling 17.3K hours. We provide speech recognition (ASR) baselines and validate the versatility of VoxPopuli unlabeled data in semisupervised ASR and speech-to-text translation under challenging out-of-domain settings. The corpus is available at https://github. com/facebookresearch/voxpopuli.

Domaines: Linguistique; Sciences cognitives; Informatique; Machine Learning [stat.ML]

Changhan Wang, Morgane Rivière, Ann Lee, Anne Wu, Chaitanya Talnikar, et al.. VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation. ACL 2021 - 59th Annual Meeting of the Association for Computational Linguistics, Aug 2021, Bangkok, Thailand. ⟨10.18653/v1/2021.acl-long.80⟩. ⟨hal-03831929⟩