Research On Place Retrieval Technology Based On Knowledge Graph

With the rapid development of Internet,place retrieval has become an indispensable part of daily life.However,traditional retrieval technology can not meet the semantic needs of users.Knowledge graph has been introduced into the new generation retrieval systems to improve retrieval performance.Place retrieval methods of knowledge graph are mainly divided into graph-based database retrieval and RDF-based retrieval.Graph-based database retrieval has high retrieval speed,but it can not make full use of users' spatial and temporal information to sort the results.In addition,users need to understand the underlying storage structure and query language;RDF-based retrieval method can better express semantic information.KSP(Top-K Relevant Semantic Place Retrieval)is a retrieval method with good effect on small-scale RDF digraphs,but RDF digraphs can not express complete semantic information,and can not efficiently retrieve place on large-scale RDF undirected graphs.This thesis designs a place retrieval method for large-scale RDF undirected graph data,which aims to improve the speed and accuracy of retrieval,including data preprocessing,index construction,query algorithm and result ranking.Firstly,data preprocessing is carried out and the problem model to be solved is generated.Meanwhile,semantic distance index,spatial index and spatial semantic hybrid index are constructed based on semantic and spatial information.The index is optimized to reduce the query and storage overhead of the index.In the retrieval stage,this thesis designs a two-stage query strategy including fuzzy query and accurate query,which optimizes the dynamic boundary pruning while using index information to retrieve locations,and further improves the retrieval speed.Finally,an index Top-K ranking model based on Skyline is proposed to rank the retrieval result sets.This thesis designs and implements KPR(Knowledge-Graph Place Retrieval)system based on the above theoretical basis.By building the experimental environment and platform,and using data from the open source knowledge base Yago and DBpedia to test and analyze,and compare.The experimental results show that KPR system has higher retrieval efficiency in large data background,and it reduces the cost of index construction according to the optimization method.At the same time,it improves the retrieval accuracy and the overall performance of the retrieval system.