Joint Extraction of Entities and Relations Based on a Novel Decomposition Strategy

TL;DR

This paper introduces a novel decomposition strategy for joint entity and relation extraction, improving accuracy by reducing noise and capturing semantic interdependencies, leading to state-of-the-art results on multiple datasets.

Contribution

The paper proposes a new decomposition approach that splits the joint extraction task into subtasks and sequence labeling problems, enhancing performance and reducing noise.

Findings

Outperforms previous methods by 5.2%, 5.9%, and 21.5% in F1 score.

Achieves state-of-the-art results on three public datasets.

Effectively captures semantic interdependencies between extraction steps.

Abstract

Joint extraction of entities and relations aims to detect entity pairs along with their relations using a single model. Prior work typically solves this task in the extract-then-classify or unified labeling manner. However, these methods either suffer from the redundant entity pairs, or ignore the important inner structure in the process of extracting entities and relations. To address these limitations, in this paper, we first decompose the joint extraction task into two interrelated subtasks, namely HE extraction and TER extraction. The former subtask is to distinguish all head-entities that may be involved with target relations, and the latter is to identify corresponding tail-entities and relations for each extracted head-entity. Next, these two subtasks are further deconstructed into several sequence labeling problems based on our proposed span-based tagging scheme, which are conveniently solved by a hierarchical boundary tagger and a multi-span decoding algorithm. Owing to the reasonable decomposition strategy, our model can fully capture the semantic interdependency between different steps, as well as reduce noise from irrelevant entity pairs. Experimental results show that our method outperforms previous work by 5.2%, 5.9% and 21.5% (F1 score), achieving a new state-of-the-art on three public datasets