wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations

TL;DR

wav2vec 2.0 introduces a self-supervised learning framework that leverages unlabeled speech data to produce representations capable of outperforming semi-supervised methods in speech recognition, especially with limited labeled data.

Contribution

It presents a novel contrastive learning approach with latent space masking for speech representations, significantly reducing labeled data requirements.

Findings

Achieves 1.8/3.3 WER on Librispeech with full data

Outperforms previous methods with only one hour of labeled data

Uses 53k hours of unlabeled data to reach competitive WERs with minimal labeled data

Abstract

We show for the first time that learning powerful representations from speech audio alone followed by fine-tuning on transcribed speech can outperform the best semi-supervised methods while being conceptually simpler. wav2vec 2.0 masks the speech input in the latent space and solves a contrastive task defined over a quantization of the latent representations which are jointly learned. Experiments using all labeled data of Librispeech achieve 1.8/3.3 WER on the clean/other test sets. When lowering the amount of labeled data to one hour, wav2vec 2.0 outperforms the previous state of the art on the 100 hour subset while using 100 times less labeled data. Using just ten minutes of labeled data and pre-training on 53k hours of unlabeled data still achieves 4.8/8.2 WER. This demonstrates the feasibility of speech recognition with limited amounts of labeled data.