Towards End-to-end Unsupervised Speech Recognition

04/05/2022

∙

by   Alexander H. Liu, et al.

∙

0

∙

share

Unsupervised speech recognition has shown great potential to make Automatic Speech Recognition (ASR) systems accessible to every language. However, existing methods still heavily rely on hand-crafted pre-processing. Similar to the trend of making supervised speech recognition end-to-end, we introduce  which does away with all audio-side pre-processing and improves accuracy through better architecture. In addition, we introduce an auxiliary self-supervised objective that ties model predictions back to the input. Experiments show that  improves unsupervised recognition results across different languages while being conceptually simpler.