Ask Question First for Enhancing Lifelong Language Learning

TL;DR

This paper introduces AQF-RQ, a novel approach for lifelong language learning that improves pseudo data generation and reduces catastrophic forgetting by asking questions first, achieving near multi-task performance.

Contribution

It proposes a new data format and training task that enhance pseudo question generation, improving lifelong learning in NLP with minimal performance loss.

Findings

AQF-RQ generates more accurate pseudo data.

It is robust to varying pseudo-data availability.

Performance is within 0.36% of multi-task learning.

Abstract

Lifelong language learning aims to stream learning NLP tasks while retaining knowledge of previous tasks. Previous works based on the language model and following data-free constraint approaches have explored formatting all data as "begin token (\textit{B}) + context (\textit{C}) + question (\textit{Q}) + answer (\textit{A})" for different tasks. However, they still suffer from catastrophic forgetting and are exacerbated when the previous task's pseudo data is insufficient for the following reasons: (1) The model has difficulty generating task-corresponding pseudo data, and (2) \textit{A} is prone to error when \textit{A} and \textit{C} are separated by \textit{Q} because the information of the \textit{C} is diminished before generating \textit{A}. Therefore, we propose the Ask Question First and Replay Question (AQF-RQ), including a novel data format "\textit{BQCA}" and a new training task to train pseudo questions of previous tasks. Experimental results demonstrate that AQF-RQ makes it easier for the model to generate more pseudo data that match corresponding tasks, and is more robust to both sufficient and insufficient pseudo-data when the task boundary is both clear and unclear. AQF-RQ can achieve only 0.36\% lower performance than multi-task learning.