Continual Learning Optimizations for Auto-regressive Decoder of Multilingual ASR systems

TL;DR

This paper introduces four optimization techniques for continual learning in multilingual ASR auto-regressive decoders, significantly improving performance on unseen languages without degrading existing language accuracy.

Contribution

It proposes novel decoder-specific optimizations tailored for continual learning in multilingual ASR, addressing limitations of existing methods.

Findings

Reduced average word error rate from 14.2% to 12.4% on unseen languages

Maintained performance on pre-trained languages during adaptation

Demonstrated effectiveness on Whisper model with 10 new languages

Abstract

Continual Learning (CL) involves fine-tuning pre-trained models with new data while maintaining the performance on the pre-trained data. This is particularly relevant for expanding multilingual ASR (MASR) capabilities. However, existing CL methods, mainly designed for computer vision and reinforcement learning tasks, often yield sub-optimal results when directly applied to MASR. We hypothesise that this is because CL of the auto-regressive decoder in the MASR model is difficult. To verify this, we propose four optimizations on the decoder. They include decoder-layer gradient surgery, freezing unused token embeddings, suppressing output of newly added tokens, and learning rate re-scaling. Our experiments on adapting Whisper to 10 unseen languages from the Common Voice dataset demonstrate that these optimizations reduce the Average Word Error Rate (AWER) of pretrained languages from 14.2% to 12.4% compared with Experience Replay, without compromising the AWER of new languages.