Towards combinational relation linking over knowledge graphs

Given a knowledge graph and natural language phrase, relation linking aims to find relations (predicates or properties) from the underlying match phrase. It is very useful in many applications, such as question answering, personalized recommendation text summarization. However, previous algorithms usually produce single for input phrase pay little attention more general challenging problem, i.e., combinational that extracts subgraph pattern compound (e.g. father-in-law). In this paper, we focus on task of over graphs. To resolve define several elementary meta patterns which can be used build any relation. Then design systematic method based data-driven assembly technique, performed under guidance patterns. enhance system’s understanding ability, introduce external during process. Finally, extensive experiments real confirm effectiveness proposed method.