Deriving similarity graphs from open linked data on Semantic Web

As increasing linked datasets are progressively published on the semantic Web, discovering the most similar entities in large linked datasets becomes crucial in many semantic applications. Conventional approaches usually draw upon either ontology taxonomy or relationships unilaterally. In this paper, we present a novel approach which utilizes node and link types together with the topology of semantic graph to derive a similarity graph from linked datasets. Firstly, semantic similarity transition is proposed to calculate the similarity between two resources. Furthermore, a system is developed to derive and visualize the similarity graph based on the calculated similarity. We apply this approach to a real-world linked dataset generated in healthcare domain and the evaluation result shows that our method yields a satisfying result in a use case of clinical decision-making.