A Heterogeneous Information Network Method for Entity Set Expansion in Knowledge Graph

Abstract

Entity Set Expansion (ESE) is an important data mining task, e.g. query suggestion. It aims to expand an entity seed set to obtain more entities which have traits in common. Traditionally, text and Web information are widely used for ESE. Recently, some ESE methods employ Knowledge Graph (KG) to extend entities. However, these methods usually fail to sufficiently and efficiently utilize the rich semantics contained in KG. In this paper, we use the Heterogeneous Information Network (HIN) to represent KG, which would effectively capture hidden semantic relations between seed entities. However, the complex KG introduces new challenges for HIN analysis, such as generation of meta paths between entities and addressing ambiguity caused by multiple types of objects. To solve these problems, we propose a novel Concatenated Meta Path based Entity Set Expansion method (CoMeSE). With the delicate design of the concatenated meta path generation and multi-type-constrained meta path, CoMeSE can quickly and accurately detect important path features in KG. In addition, heuristic learning and PU learning are employed to learn the weights of extracted meta paths. Extensive experiments on real dataset show that the CoMeSE accurately and quickly expands the given small entity set.