Preventing Catastrophic Forgetting in Continual Learning of New Natural Language Tasks

TL;DR

This paper introduces a distillation-based method to incrementally expand multi-task learning models in NLP, effectively preventing catastrophic forgetting and maintaining performance on old tasks while learning new ones.

Contribution

It proposes a novel knowledge distillation approach using unlabeled data to prevent forgetting in continual learning of NLP tasks, reducing retraining costs.

Findings

Prevents up to 20% performance drops on old tasks.

Effective in practical voice assistant scenarios.

Maintains high performance on new tasks.

Abstract

Multi-Task Learning (MTL) is widely-accepted in Natural Language Processing as a standard technique for learning multiple related tasks in one model. Training an MTL model requires having the training data for all tasks available at the same time. As systems usually evolve over time, (e.g., to support new functionalities), adding a new task to an existing MTL model usually requires retraining the model from scratch on all the tasks and this can be time-consuming and computationally expensive. Moreover, in some scenarios, the data used to train the original training may be no longer available, for example, due to storage or privacy concerns. In this paper, we approach the problem of incrementally expanding MTL models' capability to solve new tasks over time by distilling the knowledge of an already trained model on n tasks into a new one for solving n+1 tasks. To avoid catastrophic forgetting, we propose to exploit unlabeled data from the same distributions of the old tasks. Our experiments on publicly available benchmarks show that such a technique dramatically benefits the distillation by preserving the already acquired knowledge (i.e., preventing up to 20% performance drops on old tasks) while obtaining good performance on the incrementally added tasks. Further, we also show that our approach is beneficial in practical settings by using data from a leading voice assistant.