Pre-trained Language Model Based Active Learning for Sentence Matching
Computation and Language
2020-10-13 v1
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.
Cite
@article{arxiv.2010.05522,
title = {Pre-trained Language Model Based Active Learning for Sentence Matching},
author = {Guirong Bai and Shizhu He and Kang Liu and Jun Zhao and Zaiqing Nie},
journal= {arXiv preprint arXiv:2010.05522},
year = {2020}
}
Comments
Accepted by the conference of coling 2020