Research On Semantic Similarity Measure Method For RDF Graphs

With the rapid increase of data on the Internet, shortages of keyword retrieval emerge gradually. Keyword retrieval can only provide limited structural query and lost semantic information, in addition, keyword retrieval lacks the ablility of query optimization, as a result, quantity of retrieval results is great. It is more and more difficult for users to get the required information from the Internet.Object-Level information retrieval, which has drawn more and more attention and research, can solve the above problems effectively. Object-Level information retrieval models relation data in object perspective, and describes the semantic information intuitively, thus its retrieval results are objects that contains complete entity information which can be understood easily, instead of text containing the keyword.In order to realize Object-Level information retrieval, a tool for representing the object is essential. This thesis chooses RDF(Resource Description Framework) to describe objects. RDF recommended by the W3C(World Wide Web Consortium) is the standard of describing the metadata on the World Wide Web, so more and more information is expressed by RDF.This thesis studys dominated methods of semantic similarity measure between objects represented by RDF graphs, at the same time analyzes the advandages and disadvantages of the methods, then proposes the SaS (Structure and Semantics) method. SaS perfects the algorthm of semantic similarity of RDF graphs of the Radow Oldakowski method(RO), and reflects the structural information of RDF graphs and calculates semantic similarity of concept nodes by combining improved semantic distance formulas based on the network model and weight measure method based on the information quantity model. SaS can distinguish effectively concept nodes that the above models can’t because it considers both depth and density, therefore the accuracy of semantic discrimination is raised.To verify effectiveness and feasibility of SaS, prototype system is designed and realized in the field of computer programmers recruitment. The experimental results reveal that the precision of the SaS method is better than that of RO.