Incorporating Text into the Triple Context for Knowledge Graph Embedding

Knowledge graph embedding, aiming to represent entities and relations in a knowledge graph as low-dimensional real-value vectors, has attracted the attention of a large number of researchers. However, most of the embedding methods ignore the incompleteness of the knowledge graphs and they focus on the triples themselves in the knowledge graphs. In this paper, we try to introduce the information of texts to enhance the performances based on contextual model for knowledge graph embedding. Based on the assumption of the distant supervision, the sentences in texts contains abundant semantic information of the triples in knowledge graph, so that these semantic information can be utilized to relief the incompleteness of knowledge graphs and enhance the performances of knowledge graph embedding. Compared with state-ofthe-art systems, preliminary evaluation results show that our proposed method obtains the better results in Hits@10.