GSSC: Graph Summarization based on both Structure and Concepts

In this paper, we propose a new method for graph summarization named GSSC, Graph Summarization based on both Structure and Concepts. In this method, an attributed graph is summarized by considering both of its topology and related concepts. In this method, for a given attributed graph a new graph is constructed that an edge in this new graph represents structural and conceptual similarity of its two end points. Structural and conceptual similarity of two nodes not necessarily has the equal amount of importance in the weight of the resulting edge. For example, for a special case such as query answering, structure can be more important and vice versa. Similarity of two nodes is computed based on Jaccard similarity. This method has some advantages such as flexibility, simplicity, learning capability, user-orientation that makes it a better method for graph summarization. We implemented our method and the method proposed by Bei and evaluated these two methods on real-life dataset HEP_TH. Our experimental results showed effectiveness and efficiency of our proposed method.