Multiview Translation Learning for Knowledge Graph Embedding

Figure 2

A multiview knowledge learning process of the MvTransE model. The blue arrow indicates a subgraph generation operation for extracting local relational facts from the knowledge graph; and the white arrow indicates a graph embedding process projecting the original knowledge graph and generated subgraphs into a global-view vector space and several local-view vector spaces, respectively. In each vector space, the colored coordinates represent different vector spaces. The solid circles represent entity vectors in a vector space whose color is the same as the color of the space coordinate. The four-angle stars represent relation vectors, and its color corresponds to the color of relations in the knowledge graph.