More Speaking or More Speakers?
This work addresses the problem of optimizing dataset composition for ASR researchers, but it is incremental as it analyzes existing methods without introducing new algorithms.
The paper analyzes how the composition of labeled and unlabeled datasets affects self-training and self-supervised learning in automatic speech recognition, finding that SSL needs large unlabeled data for high accuracy, while ST requires sufficient speakers in labeled data, especially in low-data regimes.
Self-training (ST) and self-supervised learning (SSL) methods have demonstrated strong improvements in automatic speech recognition (ASR). In spite of these advances, to the best of our knowledge, there is no analysis of how the composition of the labelled and unlabelled datasets used in these methods affects the results. In this work we aim to analyse the effect of number of speakers in the training data on a recent SSL algorithm (wav2vec 2.0), and a recent ST algorithm (slimIPL). We perform a systematic analysis on both labeled and unlabeled data by varying the number of speakers while keeping the number of hours fixed and vice versa. Our findings suggest that SSL requires a large amount of unlabeled data to produce high accuracy results, while ST requires a sufficient number of speakers in the labelled data, especially in the low-regime setting. In this manner these two approaches improve supervised learning in different regimes of data composition.