Systematic Analysis of Joint Entity and Relation Extraction Models in Identifying Overlapping Relations

Abstract

Named entity recognition and relation extraction are two fundamental tasks in the domain of natural language processing. Joint entity and relation extraction models have attracted more and more attention due to the performance advantage. However, there are difficulties in identifying overlapping relations among the models. To investigate the differences of structures and performances of joint extraction models, this paper implements a list of state-of-the-art joint extraction models and compares their difference in identifying overlapping relations. Experiment results show that the models by separating entity features and relation features work better than the models with feature fusion in identifying overlapping relations on three publicly available datasets.