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

TLDR

wav2vec 2.0 learns speech representations by masking latent features and solving a contrastive task over their quantized embeddings. The method attains state‑of‑the‑art WERs—1.8/3.3 on Librispeech with full data, 4.8/8.2 with only ten minutes of labeled data—and outperforms prior semi‑supervised approaches, demonstrating that accurate speech recognition is achievable 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.