Research Of Federated Query Method For Linked Data Based On Semi-join

Motivated by the booming of the semantic web and the growing amount of semantic data sources available on the Web,new challenges to query information from the mass sources with high effective and accuracy are emerging.Linked data as the form of publishing the information about the semantic web is the foundation to realize large-scale applications of the semantic web.Therefore,the research on the linked data query directly affects the development of the semantic web.Classified by the form of data source store,the federated query to SPARQL Endpoints are selected as research point.The federated query to the linked data requires joining data provided by multiple sources.Therefore,sophisticated optimization techniques are necessary for efficient query processing.In this article,the optimization method based on semi-join for the federated query to the linked data is proposed,which is from the aspect of reduce the transmission cost of intermediate results.Given the processing step and the characteristics of the linked data,the data source selection algorithm based on the voiD and ASK query is presented to improve the accuracy of the data source selection.To estimate the size of intermediate result,the estimation method oriented each basic graph pattern is raised.Besides,an algorithm for guiding the query rewritten is put forward so as to realize the mapping between the triple patterns and SPARQL Endpoints.After a query is preprocessed by the above algorithms,the core method which executing query through semi-join is presented.And in order to reduce the waiting time during query process,the idea transmitting intermediate results through RDF stream is introduced.Finally,the parallelized semi-join operation method is discussed after the analyzing of the cost model and the optimization objective.The result of the experiment show that the optimization method based on semi-join for the federated query can reduce the size of intermediate results and the time of query processing significantly and the conclusion improved query efficiency can be drawn.