CHNE: Context-aware Heterogeneous Network Embedding

A heterogeneous information network is a network with different kinds of nodes and edges, which can express different interrelationships between objects in a single graph. If we use heterogeneous information networks for analysis, we can take advantage of more information for analysis than homogeneous information networks, thus improve the quality of analysis results. Recently, many embedding processes are used to vectorize network data so that they can be used for analysis. Network embedding is a very important process in the process of network analysis, with the aim of producing vectors that express network information well, such as network structure and node features. There have been several recent network embedding studies which enhance the embedding performance by considering both network structure and text features of nodes in a network. But they are difficult to apply to heterogeneous information networks, because their network structures are different from each other. In this paper, therefore, we propose CHNE, a new heterogeneous information network embedding model which takes advantage of all interrelationships, text information and node types of heterogeneous information networks.