Three Sentences Are All You Need: Local Path Enhanced Document Relation Extraction

TL;DR

This paper introduces a simple heuristic method to select key evidence sentences for document-level relation extraction, achieving superior performance over complex graph neural network methods by combining it with BiLSTM.

Contribution

The paper proposes an effective and straightforward heuristic for evidence sentence selection in document RE, improving performance without complex models.

Findings

Outperforms graph neural network-based methods on benchmarks

Simple heuristic combined with BiLSTM yields high accuracy

Code released for reproducibility

Abstract

Document-level Relation Extraction (RE) is a more challenging task than sentence RE as it often requires reasoning over multiple sentences. Yet, human annotators usually use a small number of sentences to identify the relationship between a given entity pair. In this paper, we present an embarrassingly simple but effective method to heuristically select evidence sentences for document-level RE, which can be easily combined with BiLSTM to achieve good performance on benchmark datasets, even better than fancy graph neural network based methods. We have released our code at https://github.com/AndrewZhe/Three-Sentences-Are-All-You-Need.