Using Adapters to Overcome Catastrophic Forgetting in End-to-End Automatic Speech Recognition

Abstract

Learning a set of tasks in sequence remains a challenge for artificial neural networks, which, in such scenarios, tend to suffer from Catastrophic Forgetting (CF). The same applies to End-to-End (E2E) Automatic Speech Recognition (ASR) models, even for monolingual tasks. In this paper, we aim to overcome CF for E2E ASR by inserting adapters, small architectures of few parameters which allow a general model to be fine-tuned to a specific task, into our model. We make these adapters task-specific, while regularizing the parameters of the model shared by all tasks, thus stimulating the model to fully exploit the adapters while keeping the shared parameters to work well for all tasks. Our method outperforms all baselines on two monolingual experiments while being more storage efficient and without requiring the storage of data from previous tasks.

Accepted at ICASSP 2023. 5 pages

References in corpus (4)

• PathNet: Evolution Channels Gradient Descent in Super Neural Networks
• Continual Lifelong Learning in Natural Language Processing: A Survey
• Layer-wise Fast Adaptation for End-to-End Multi-Accent Speech Recognition
• Continual Learning for Monolingual End-to-End Automatic Speech Recognition

Cited by in corpus (5)

• Learning from models beyond fine-tuning
• Weight Averaging: A Simple Yet Effective Method to Overcome Catastrophic Forgetting in Automatic Speech Recognition
• Unsupervised Accent Adaptation Through Masked Language Model Correction Of Discrete Self-Supervised Speech Units
• Continual Learning With Quasi-Newton Methods
• Inverse-Hessian Regularization for Continual Learning in ASR