Unpaired text and audio injection have emerged as dominant methods for improving ASR performance in the absence of a large labeled corpus. However, little guidance exists on deploying these methods to improve production ASR systems that are trained on very large supervised corpora and with realistic requirements like a constrained model size and CPU budget, streaming capability, and a rich lattice for rescoring and for downstream NLU tasks. In this work, we compare three state-of-the-art semi-supervised methods encompassing both unpaired text and audio as well as several of their combinations in a controlled setting using joint training. We find that in our setting these methods offer many improvements beyond raw WER, including substantial gains in tail-word WER, decoder computation during inference, and lattice density.
@article{arxiv.2304.11053,
title = {A Comparison of Semi-Supervised Learning Techniques for Streaming ASR at Scale},
author = {Cal Peyser and Michael Picheny and Kyunghyun Cho and Rohit Prabhavalkar and Ronny Huang and Tara Sainath},
journal= {arXiv preprint arXiv:2304.11053},
year = {2023}
}