Integrating Relation Constraints with Neural Relation Extractors

TL;DR

This paper introduces a novel framework that incorporates relation constraints into neural relation extraction models using a new loss term, improving consistency and accuracy across multiple datasets.

Contribution

It proposes a unified approach with ConstraintLoss to integrate relation constraints into neural models, enhancing their performance and consistency.

Findings

Outperforms existing neural relation extraction models on English and Chinese datasets.

Reduces inconsistency among local predictions by incorporating relation constraints.

Demonstrates effectiveness of the approach even without extra post-processing.

Abstract

Recent years have seen rapid progress in identifying predefined relationship between entity pairs using neural networks NNs. However, such models often make predictions for each entity pair individually, thus often fail to solve the inconsistency among different predictions, which can be characterized by discrete relation constraints. These constraints are often defined over combinations of entity-relation-entity triples, since there often lack of explicitly well-defined type and cardinality requirements for the relations. In this paper, we propose a unified framework to integrate relation constraints with NNs by introducing a new loss term, ConstraintLoss. Particularly, we develop two efficient methods to capture how well the local predictions from multiple instance pairs satisfy the relation constraints. Experiments on both English and Chinese datasets show that our approach can help NNs learn from discrete relation constraints to reduce inconsistency among local predictions, and outperform popular neural relation extraction NRE models even enhanced with extra post-processing. Our source code and datasets will be released at https://github.com/PKUYeYuan/Constraint-Loss-AAAI-2020.