Schema-Rich Heterogeneous Network Mining

Abstract

Traditional heterogeneous information network usually has simple network schema, where there are a small number of types of nodes and links and meta paths are easily enumerated. However, in many real applications, some heterogeneous information networks have a huge number of types of nodes and links, and it is hard to depict their network schema. We call this kind of networks as schema-rich heterogeneous information network. For example, knowledge graph, constructed with \( \) tuples, can be considered as a schema-rich heterogeneous network, where there are usually tens of thousands of types of nodes and links. In this chapter, we introduce two data mining tasks on schema-rich heterogeneous network: link prediction and entity set expansion. Through these two tasks, we illustrate the challenges and potential solutions on mining this kind of more complex and popular heterogeneous networks.