Joint Entity Relation Extraction based on Graph Neural Network

Abstract

In the task of joint entity relation extraction, the problem of redundant relations caused by multiple types of relation extraction in overlapping entities and the problem of overlapping triples remain many challenging problems. To address the above issues, we propose a joint entity relation extraction model based on attention mechanism and multi-layer graph convolutional network (att-MLGCN) that fuses semantic and dependency information. We combine not only semantic features but also syntactic dependency features to fully extract entity features, predict various relation types between entity pairs. We conduct experiments on two widely used public datasets, NYT and WebNLG. The results show that our model has some improvement over the baseline model and some models.