Pre-trained Language Model Based Active Learning for Sentence Matching

TL;DR

This paper introduces a novel active learning method for sentence matching that leverages pre-trained language models to select more informative instances, reducing annotation costs and improving accuracy.

Contribution

It proposes a pre-trained language model based active learning approach that incorporates linguistic criteria for more effective instance selection in sentence matching.

Findings

Achieves higher accuracy with fewer labeled instances

Outperforms entropy-based active learning methods

Demonstrates efficiency in annotation process

Abstract

Active learning is able to significantly reduce the annotation cost for data-driven techniques. However, previous active learning approaches for natural language processing mainly depend on the entropy-based uncertainty criterion, and ignore the characteristics of natural language. In this paper, we propose a pre-trained language model based active learning approach for sentence matching. Differing from previous active learning, it can provide linguistic criteria to measure instances and help select more efficient instances for annotation. Experiments demonstrate our approach can achieve greater accuracy with fewer labeled training instances.