Large-scale ASR Domain Adaptation using Self- and Semi-supervised Learning
This work addresses domain adaptation for online ASR models, which is incremental as it applies existing self- and semi-supervised learning methods to a new large-scale scenario.
The paper tackled the problem of adapting automatic speech recognition (ASR) models to unseen domains in a large-scale production setting, achieving a 13.5% relative word error rate (WER) improvement for the target domain using only 3% of its data compared to a full data baseline.
Self- and semi-supervised learning methods have been actively investigated to reduce labeled training data or enhance the model performance. However, the approach mostly focus on in-domain performance for public datasets. In this study, we utilize the combination of self- and semi-supervised learning methods to solve unseen domain adaptation problem in a large-scale production setting for online ASR model. This approach demonstrates that using the source domain data with a small fraction of the target domain data (3%) can recover the performance gap compared to a full data baseline: relative 13.5% WER improvement for target domain data.