Simple BERT Models for Relation Extraction and Semantic Role Labeling

TL;DR

This paper demonstrates that simple BERT-based models, without external features, can achieve state-of-the-art results in relation extraction and semantic role labeling, establishing strong baselines for future research.

Contribution

First to successfully apply BERT directly to relation extraction and semantic role labeling without external features, achieving state-of-the-art performance.

Findings

BERT-based models outperform previous methods.

External features are unnecessary for high performance.

Models serve as strong baselines for future work.

Abstract

We present simple BERT-based models for relation extraction and semantic role labeling. In recent years, state-of-the-art performance has been achieved using neural models by incorporating lexical and syntactic features such as part-of-speech tags and dependency trees. In this paper, extensive experiments on datasets for these two tasks show that without using any external features, a simple BERT-based model can achieve state-of-the-art performance. To our knowledge, we are the first to successfully apply BERT in this manner. Our models provide strong baselines for future research.