Research On Correlation Of Intelligence Based On Knowledge Graphs

Knowledge Graphs is an emerging technology put forward in recent years.Once its concept is introduced,it has received widespread attention in academia and industry.Because knowledge Graphs can well represent the correlation between knowledge,it is also an inevitable and important practical application in the field of intelligence correlation analysis.In recent years,for the study of knowledge atlas,knowledge mining and retrieval methods are the hotspots of research.Especially in the context of big data,the number of nodes and relationships in the knowledge Graphs also grow exponentially.How to construct massive knowledge Graphs and excavation on large-scale knowledge Graphs has always been a difficult problem to be solved.Although many models and methods for knowledge Graphs are proposed by some scholars,but they all have some problems such as high computational complexity and poor predictive effect.Therefore,how to effectively excavate and analyze in mass knowledge Graphs is the key issue of current research.In this thesis,two major tasks of relational reasoning and correlation analysis in intelligence knowledge Graphs are proposed,and improved relational reasoning and relational inquiring methods are proposed.In terms of relational reasoning,aiming at the shortcomings of traditional methods in multi-relational prediction,the undirected graph probability method with two-way semantics to measure the reliability of the relational path is adopted in this thesis.And constructs the relation reasoning method with the embedded model TransE.In relational query,this thesis presents a novel RDF knowledge graphs correlation retrieval method based on extension.The idea of this method is expanding the expression of the original RDF by using keywords and weights,and to consider the influence of data classification on the diversity of results in the calculation of weights.Based on the extended RDF knowledge Graphs,a query relaxation model of three-tuples similarity is proposed.The similarity calculation is based on a hybrid evaluation model of semantic overlap and graph embedding.Finally,the query result set is sorted by the weight of the triple and the keyword,and the top-k query results with relevance are returned.The experimental results show that the improved relational reasoning model proposed in this paper has a marked improvement on the prediction accuracy in the multi-relational prediction compared with the previous methods.In the relevance of inquiries,the experimental results show that the recall rate and accuracy have a very good improvement.Compared with the traditional SPARQL query,the proposed method overcomes the restriction of query conditions,and the query result set has strong correlation and diversity.