Improving Temporal Generalization of Pre-trained Language Models with Lexical Semantic Change

TL;DR

This paper introduces a lexical-level masking strategy to improve the temporal generalization of pre-trained language models by addressing lexical semantic change, outperforming traditional continual training methods.

Contribution

The paper proposes a novel lexical-level masking approach for post-training language models to enhance their ability to adapt over time, focusing on lexical semantic change.

Findings

The proposed method improves temporal generalization on multiple datasets.

It outperforms existing continual training approaches.

The approach is effective across different models and tasks.

Abstract

Recent research has revealed that neural language models at scale suffer from poor temporal generalization capability, i.e., the language model pre-trained on static data from past years performs worse over time on emerging data. Existing methods mainly perform continual training to mitigate such a misalignment. While effective to some extent but is far from being addressed on both the language modeling and downstream tasks. In this paper, we empirically observe that temporal generalization is closely affiliated with lexical semantic change, which is one of the essential phenomena of natural languages. Based on this observation, we propose a simple yet effective lexical-level masking strategy to post-train a converged language model. Experiments on two pre-trained language models, two different classification tasks, and four benchmark datasets demonstrate the effectiveness of our proposed method over existing temporal adaptation methods, i.e., continual training with new data. Our code is available at \url{https://github.com/zhaochen0110/LMLM}.