vq-wav2vec: Self-Supervised Learning of Discrete Speech Representations

Code

• github.com/eastonYi/wav2vec
  pytorch★ 170
• github.com/clovaai/textual-kd-slu
  pytorch★ 8

Abstract

We propose vq-wav2vec to learn discrete representations of audio segments through a wav2vec-style self-supervised context prediction task. The algorithm uses either a gumbel softmax or online k-means clustering to quantize the dense representations. Discretization enables the direct application of algorithms from the NLP community which require discrete inputs. Experiments show that BERT pre-training achieves a new state of the art on TIMIT phoneme classification and WSJ speech recognition.

Tasks

Benchmark Results

Reproductions