Text-Enhanced Question Answering over Knowledge Graph

Question answering over knowledge graph is an important area of research within question answering. Existing methods mainly focus on the utilization of information in knowledge graphs and ignore the abundant external information of entities. However, knowledge graphs are usually incomplete and entities in knowledge graphs are not completely described. In this paper, we propose a novel text-enhanced question answering model over knowledge graph by taking advantage of the rich context information in a text corpus. We believe the rich textual context information can effectively alleviate the information loss in knowledge graphs and enhance the knowledge representation capability in the answer end. To this end, we apply an attention model to realize dynamic fusion of internal and external information. Besides, Transformer Encoder network is used to obtain the representation of input question and descriptive text. The experiments on the WebQuestions dataset prove that compared with other state-of-the-art QA methods, our method can effectively improve the accuracy.