Algorithm Optimization And System Implementation Of Knowledge Graph Relational Reasoning Based On Deep Learning

As an effective way of expressing information,knowledge graphs have been widely used in scenarios such as knowledge search and robot question answering.However,there are still serious data quality and data missing problems in current knowledge graphs.In order to solve the above problems,knowledge graph relational reasoning uses known entity relationship information to reason about the potential relationship between entities by analyzing and computing the knowledge in the graph.This thesis takes knowledge graph relational reasoning as the main research object,proposes to use graph neural network to improve the information modeling process of nodes in the graph,applies the reasoning method to the military field,and designs a reasoning service platform based on military knowledge.The main content of the thesis is as follows:1.Aiming at the problem of insufficient utilization of node structure information in the process of knowledge reasoning,a graph neural network is introduced to enhance the information of nodes in the graph.This thesis fully considers the node’s own representation and the distribution of surrounding neighbors,uses the Graph Attention Network(Graph Attention Network,GAT)to capture the interaction characteristics of different intensities between the central entity node and the surrounding nodes,and aggregates the pointing information of the neighbors within k hops of the node,so that the network is able to learn implicit semantic associations between entities.The improved model was tested on the benchmark data sets WN18RR,FB15K-237,and NELL-995,and the results showed that the method effectively improved the inference performance.2.Aiming at the problems of low quality of military knowledge graph data and inaccurate knowledge representation,a knowledge graph relational reasoning system in the military field is designed and implemented.This thesis collects and organizes public military data,constructs a knowledge graph in the military field,and uses the knowledge graph relational reasoning algorithm to reason and complete the missing military knowledge in the graph.The functions of the system mainly include graph visualization,knowledge reasoning prediction and knowledge quiz,etc.To sum up,this thesis proposes a knowledge graph relational reasoning method that integrates entity node information,and implements a relational reasoning system in the military field,providing a new method and approach for people to better understand and apply military knowledge.