Multi-knowledge Graph Fusion And Fast Retrieval Method Based On Social Network

The development of knowledge graph technology has promoted the efficient organization of massive data.Many experts and institutions have created various knowledge graphs,but there exists a problem of different semantic expressions for the same entity due to their varying sources of information and knowledge description systems,which is unfavorable for mining and utilizing knowledge within knowledge graphs.To solve these problems,this paper focuses on the research of multi-knowledge graph integration and rapid retrieval methods and conducts practical experiments based on a social network knowledge graph.The main work is as follows:(1)This paper investigates and analyzes the current research status and relevant technological background of entity alignment,knowledge reasoning,and knowledge graph retrieval technology,revealing the entity alignment problem and relationship loss problem in multi-knowledge graph integration.(2)Based on the above problems and content,two solutions are proposed in this paper.The first solution is an entity alignment method based on neighborhood matching.In addition to comparing the similarity of adjacent entities based on string and semantic similarity,this method also utilizes semantic information of linking relationships and mapping properties for relationship matching.Through continuous iteration of entity alignment and relationship alignment,the accuracy of entity alignment can be enhanced.The second solution is a relationship prediction method based on Monte Carlo tree search,which combines explicit reasoning with implicit prediction.In the process of explicit reasoning,rule induction and sequential decisionmaking are achieved using the Monte Carlo tree search method.Implicit prediction compensates for the incompleteness of explicit reasoning through embedding learning.(3)To verify and evaluate the above two methods,cross-language dataset and social relationship dataset are used in the experiments.The results show that compared with traditional entity alignment methods,the entity alignment method based on neighborhood matching proposed in this paper has higher accuracy.Compared with traditional relationship prediction methods,the Monte Carlo tree search-based relationship prediction method proposed in this paper has better interpretability and prediction accuracy.(4)This paper constructs a knowledge graph question-answering system based on a social network.By constructing an efficient index,the system enables fast retrieval and utilization of the knowledge graph.In addition,the relationship prediction method based on Monte Carlo tree search is also applied to the knowledge graph questionanswering system,expanding the scope of the knowledge graph retrieval and achieving missing relationship retrieval based on relationship prediction.