Entity relationship extraction model based on entity information enhancement

The entity-relationship extraction task consists of name entity recognition and relationship classification. To address the problem that the current entity-relationship extraction task in the Chinese language domain does not fully utilized the relevant information of entities, we propose a relationship extraction model based on entity information enhancement, using entity category information to construct entity category labels to enrich the original text data, adding the labeled vocabulary as a priori information to the word list of the pre-trained model. At the same time, entity information is considered a kind of crucial information in the relationship extraction task, so this paper increases the entity information in the input data to enhance the model’s ability to process the critical information of the task. Two sub-experiments, entity recognition and relationship classification, are constructed in the form of a Pipeline, where the relationship classification task exists as a downstream task of entity recognition. The model is trained and tested on the DuIE dataset. The results demonstrate the superiority of the proposed model for entity-relationship in the Chinese domain.