Research On Multi-source Data Graph Representation And Retrieval Optimization Method Oriented To Knowledge Graph

The data used in the generation of the knowledge graph has characteristics such as different sources,diverse types,and low utilization.These characteristics are likely to cause the problem of difficult characterization of multi-source data and low efficiency of knowledge graph retrieval.In view of the above problems,there is an urgent need for effective characterization and retrieval optimization methods for multisource heterogeneous data.This dissertation focuses on the research of multi-source heterogeneous text data representation and retrieval optimization methods.The main work is as follows:First,this article explores the relationship between multi-source data,knowledge graph and graph retrieval technology,and reveals the problem of multi-source data graph representation in knowledge graph and the problem of fast retrieval of multisource data under graph representation.Investigate and analyze the current status of domestic and foreign research on the issue.Second,based on the above problems and content,this paper proposes two graph representation methods.One is a bipartite graph representation method based on Markov decision process.This method is based on Markov decision process and bipartite graph concept to refine and represent multi-source The data relationship in the data can realize the fusion characterization of multi-source data,but this method has the problem of low retrieval efficiency.Based on this,this paper proposes a second representation method-a two-part bipartite graph representation method based on graph similarity measurement.This method uses the best decision algorithm and graph similarity measurement algorithm.On the basis of the first method,the original method is refined.The top-k structure in the bipartite graph forms a top-k bipartite graph,and forms a two-part bipartite graph structure with the original bipartite graph.In this structure,the top-k bipartite graph can narrow the search scope and improve the retrieval efficiency,and the two-part bipartite graph structure can realize the dynamic evolution of the graph.Third,this article uses questionnaire medical data and electronic medical record medical data to conduct experimental verification and performance evaluation of the two methods.The evaluation results show that compared with the traditional vector representation method,the bipartite graph representation method in this paper has better interpretability and characterization effect;compared with the bipartite graph representation method,the construction time of the bipartite graph representation method is reduced by 65.8% on average,The memory consumption is reduced by an average of 50.2%,and the matching time is reduced by an average of 53.4%.After the dynamic evolution of the two-part bipartite graph structure,the retrieval accuracy increased by an average of 8%.