| RDF(Resource Description Framework)is currently an important content of semantic Web research,which can make the expression of network data more standardized.The large RDF knowledge bases contain the semantics that describe the spatial location.Spatial location can be described as semantic place,and the retrieval of semantic place in the Semantic Web can improve the accuracy of spatial location retrieval.Therefore,RDF is very suitable for retrieving spatial locations.Keyword retrieval on the RDF knowledge base does not require the user to master the query language and understand the relationship between the data.Only the user can enter the keyword to find the subgraph containing the keyword.KSP(Top-k Relevant Semantic Place)retrieval is a semantic place retrieval method on RDF.It combines keyword technology and spatial location retrieval technology on RDF knowledge base.Keyword retrieval techniques are used on the RDF knowledge base to retrieve spatial location semantics.The goal of KSP retrieval is to construct the RDF knowledge base into an RDF directed graph,and find the smallest subgraph(qualified semantic place)that satisfies all keywords and roots at the location node on the RDF directed graph.The BSP(Basic Semantic Place retrieval algorithm)algorithm can implement KSP retrieval,but can only retrieve the labels on the vertices,and the edge attributes between the vertices cannot be retrieved.In view of the problem that the BSP algorithm can not retrieve the edge attribute,this paper improves the BSP algorithm.Then use the Spark framework to parallelize the improved algorithm,and do the following work:(1)The semantic search algorithm VEL-BSP(BSP based on Vertex and Edge Label)based on vertex and edge label is proposed,which can retrieve the vertex and edge labels on the RDF graph.First,find the qualified position semantics with the smallest number of vertices(optimal qualified position semantics),and the effective vertex ratio of the optimal qualified position semantics as the measure of the degree of correlation.The score function is defined by combining the effective vertex ratio,the position vertex and the query position.Rank and return Top-k position vertices based on the score function values of the best qualified position semantics.Finally,the pruning algorithm VELP-BSP(VEL-BSP with Pruning)is designed according to the dynamic range of the effective vertex ratio,and some vertex that cannot enter the Top-k is pruned.(2)In large-scale RDF data,the Spark framework is used to realize the parallelization calculation of VEL-BSP algorithm.The segmentation strategy of the position subgraph is proposed,and the RDF graph is divided into multiple position subgraphs by point segmentation.Then use HDFS to realize the distributed storage of RDF graphs.Based on this,the GraphX graph computing framework is used to parallelize the RDF graphs.Finally,the parallel design of VEL-BSP algorithm is run in Spark cluster,which improves the efficiency of semantic place retrieval in a large number of RDF knowledge bases.(3)A situational application of VEL-BSP algorithm is designed,and a hospital intelligent retrieval system is designed based on relevant hospital data.The VEL-BSP algorithm based on Spark parallelization is applied to the system.Firstly,the storage method of hospital spatial location is introduced,and then the construction process of hospital RDF knowledge base is introduced.Then the hospital intelligent retrieval system is designed,and the display of the system is given. |
