OFNE: a framework of opinion features regulated network embedding

Abstract

Network embedding technologies that transform the nodes of a network into a low-dimensional vector space have many various potential applications such as node classification, community detection and Internet public opinion analysis and so forth. Most existing approaches for network embedding are calculated by utilizing the topology structure of a target network which simply describes the social relations among nodes. However, another factor derived from the significant “opinions” is usually neglected by those works. In particular, there exist dramatic cases of Internet Public Affair, where many users with the same opinion have no social connections. On top of that, the social network can be very sparse, which is unsuitable for network embedding. Therefore, this paper proposes an efficient approach opinion feature network embedding (OFNE) that combines both social relations and opinion features within a network. OFNE defines an opinion feature edge to preserve the opinion features. Experimental results on real-world datasets across different domains demonstrate that the proposed approach OFNE outperforms the regular social network embedding approaches, especially when the opinions have explicit sentiment orientation.