Research On Semantic Similarity Calculation Of Linked Data

As the semantic web technology matures, the network resource environmentdevelops in the direction of intelligence and structuration. In current network resourceenvironment, the most of the information in this network is present in the form thatpeople can understand, not to be understood and processed by computer. Thisimpedes the development of network resource environment. The proposing ofsemantic web can solve these problems well. By adding the formal semanticinformation to the documents on the World Wide Web, semantic web can make thecomputer understand and deal with these documents to realize the automation of dataprocessing, and improve the efficiency of information searching.The appearance of linked data brings the substantial development of semanticweb and is recommended as the best practice of semantic web by W3C. By thestructural description and links between the data, Linked data is associated with thedata of scattered field to form the global giant data space that is data network. Thisdata network offers the protection to the maximum share, reusing and release of data,and offers the new opportunity to the knowledge discovery. With the increasingnumber of linked data in the data network, how to use the characteristics of linkeddata to carry on the knowledge discovery become the key issue to be solved in thecurrent study.This paper mainly does research into the semantic similarity of knowledgediscovery research. By researching and analyzing the existing semantic similaritycalculation methods about RDF data, the semantic similarity calculation method oflinked data based on multiple factors is presented. This method analyzed emphaticallythe three important factors influencing the similarity calculation, such as types ofattribute value, attribute weight and correlation, and for each of influence factorscorresponding similarity calculation formula is given. Finally the method is verifiedthrough specific instances. The experimental results show that, this method this paperproposed makes full use of the semantic information between concepts, and calculation results can better reveal the similarity relation between concepts in linkeddata.In the different application, the importance of attributes is different, and theweights of attributes can be changed. This paper carries on further researches aboutthe calculation problem of attributes weights, and presents a semantic similaritycalculation method of linked data based on dynamic weight. This method improvesthe semantic similarity calculation method proposed by Song D based on the Tverskycalculation model. First, this method dynamically computes the attribute weightaccording to quantity of different attribute values, distribution of attribute values, andvalidity of attribute. This method can distinguish the instances well that methodproposed by Song D and Tversky calculation model cannot distinguish, and improvethe accuracy of semantic similarity calculation about Linked data. Finally, in order toverify the validity and stability of this method proposed in this paper, ACM andFOAF testing datasets are used to experiment. Comparing to existing sentencesimilarity computing methods, the result of experiment shows that the methodproposed in this paper is superior to other similarity calculation methods in thevalidity and stability.