Alexei Baevski, Steffen Schneider, Michael Auli
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.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Speech Recognition | TIMIT | Percentage error | 11.6 | vq-wav2vec |