We revisit self-training in the context of end-to-end speech recognition. We demonstrate that training with pseudo-labels can substantially improve the accuracy of a baseline model. Key to our approach are a strong baseline acoustic and language model used to generate the pseudo-labels, filtering mechanisms tailored to common errors from sequence-to-sequence models, and a novel ensemble approach to increase pseudo-label diversity. Experiments on the LibriSpeech corpus show that with an ensemble of four models and label filtering, self-training yields a 33.9% relative improvement in WER compared with a baseline trained on 100 hours of labelled data in the noisy speech setting. In the clean speech setting, self-training recovers 59.3% of the gap between the baseline and an oracle model, which is at least 93.8% relatively higher than what previous approaches can achieve.
@article{arxiv.1909.09116,
title = {Self-Training for End-to-End Speech Recognition},
author = {Jacob Kahn and Ann Lee and Awni Hannun},
journal= {arXiv preprint arXiv:1909.09116},
year = {2020}
}
Comments
To be published in the 45th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2020