Structural similarity computation based on extended edge matching method

Similarity measurement of hierarchically structured data is crucial in the data mining, database and information retrieval communities. In this paper, we propose an extended edge matching method for better evaluating structural similarity between XML(eXtensible Markup Language) documents. The proposed method not only generates edges between parent nodes and child nodes, but also generates topological edges between ancestor nodes and descendant nodes. Furthermore, complete, topological and repeated matchings are distinguished in the process of edge matching. When one edge matches another, the method can identify the type of matching, and assign proper weight to it. Experiments demonstrated that the proposed method generated better similarity results and clustering results in comparison with some existing methods.