Design And Implementation A System Of Top-k Semantic Place Retrieval On Spatial RDF Graph

RDF data are accessed using structured query languages,such as SPARQL.However,this requires users to understand the language as well as the RDF schema.Keyword search on RDF data aims at relieving the user from these requirements.We designed and implemented a top-k semantic place retrieval system on spatial RDF data.The goal of top-k relevant semantic place retrieval(k SP)is to find RDF sub-graphs of spatial entities that contain query keywords and are located near the query location.The novelty of k SP queries is that they are location-aware and that they do not rely on the use of structured query language.Our system is mainly composed of four parts: user request module,data preprocessing module,query processing and result display module.The core query processing module contains an algorithm for processing k SP queries.The algorithm uses the R-tree index to find the spatial entities near the query position first.Then it goes to RDF graph and finds a sub-tree containing all query keywords with each spatial entity as the root node.The similarity between semantic place and the query is calculated according to the distance from the spatial entity to the query position and the looseness the corresponding sub-tree.Finally the top-k semantic positions are selected as the result.In order to improve the performance of the system to deal with large-scale queries,we designed and compared two parallel query processing algorithms: Basic Parallel Semantic Place Retrieval Algorithm(BPSP)and Fine-grained Parallel Semantic Place Retrieval Algorithm(FGPSP).BPSP algorithm constructs a thread pool.Each thread processes a query.Idle threads automatically get the query in the queue and calculate the result.The FGPSP algorithm constructs two thread pools.The threads in the first thread pool are responsible for finding the spatial entities near the query position,and the threads in the second thread pool are responsible for retrieving the RDF graph to find the sub-tree containing all the keywords.We tested the performance of two parallel algorithms on real large-scale data.The experimental results show that the BPSP algorithm is superior to the FGPSP algorithm for the following reasons: 1)FGPSP divides the serial algorithm into two parts.There are shared variables of this two parts,so there will be a time delay;2).Due to the limitation of systemresources,when multiple threads operate the queue at the same time,the insertion consumes much more time.