Research On Semantic Similarity Computation Method Of Linked Data Based On Multi-granularity

The Semantic Web is to solve the problem of semantic description of information,to build a new resources environment which is full of structured and semantic,intelligent information.The information can be recognized and processed by a computer.And,it realizes the communication between the person and the computer.Linked data is regarded as the best practices and the core technology of the semantic web.Linked data realizes the mutual contact and mutual connection between data by adding the semantic description of entity and building the semantic association links,which drives the construction of the Web of Linked Data.Linked data shows the incomparable advantages and potential for knowledge discovery application because of its characteristics.Linked data provides a new technology and a new resources environment for knowledge discovery.Knowledge discovery provides a new development goal and a research direction for Linked data.Therefore the knowledge discovery based on Linked data has become a research focus in semantic web.With the development and widely application of the Linked data,it shows the explosive growth trend.The research of semantic similarity calculation of Linked data must be an important research problem in order to avoid the problem of information overload and improve efficiency of finding the novel and effective knowledge from the massive Linked data resources.As a result,this paper presents the research on semantic similarity computation method of Linked Data based on multi-granularity.First,Based on analyzing the research status of Linked Data semantic similarity computation,and summarizing the existing semantic similarity calculation method,this paper analyzes the calculating thought,the function model,the applicability,advantages and disadvantages of these method.And then,it puts forward a multi-granularity Linked Data information description model based on analyzing the characteristics of Linked Data description structure.It expounds the content of the coarse,medium and fine granularity Linked Data description information.After that,this paper analyzes the influence factors of Linked Data in semantic similarity computation in detail,Which including entity attributes,the types of attribute value,the importance of attribute,the correlation distance between entities,and the correlation path between entities.On the basis of the above researches,this paper comes up with the semantic similarity computation method of Linked Data based on multi-granularity.In this method,the process of Linked Data semantic similarity calculation is divided into three modules,as the coarse-granularity description module,the medium-granularity description module,and the fine-granularity description module.Corresponding semantic similarity calculation method is given according to the features and contents of each module.Because of the different purposes,this paper presents two kinds of semantic similarity calculation process,which are for the semantic similarity computing between Linked data pairs and for obtaining similar Linked data set.Finally,this paper verifies the method by using Linked Movie Database.It analyses the experimental results,which shows that the method has the good applicability,accuracy and stability.