Continual Learning for Monolingual End-to-End Automatic Speech Recognition
This addresses the problem of enabling ASR models to adapt to new accents or topics without losing past performance, which is incremental as it applies existing continual learning methods to ASR.
The paper tackled catastrophic forgetting in monolingual end-to-end automatic speech recognition when adapting to new domains, finding that the best continual learning method reduced the performance gap by over 40% compared to joint training while using only 0.6% of the original data.
Adapting Automatic Speech Recognition (ASR) models to new domains results in a deterioration of performance on the original domain(s), a phenomenon called Catastrophic Forgetting (CF). Even monolingual ASR models cannot be extended to new accents, dialects, topics, etc. without suffering from CF, making them unable to be continually enhanced without storing all past data. Fortunately, Continual Learning (CL) methods, which aim to enable continual adaptation while overcoming CF, can be used. In this paper, we implement an extensive number of CL methods for End-to-End ASR and test and compare their ability to extend a monolingual Hybrid CTC-Transformer model across four new tasks. We find that the best performing CL method closes the gap between the fine-tuned model (lower bound) and the model trained jointly on all tasks (upper bound) by more than 40%, while requiring access to only 0.6% of the original data.