Sparse Representation Method Of Knowledge Graph Based On Boolean Matrix Decomposition

In recent years,knowledge graph,as a structured representation of human knowledge,has attracted extensive attention from academia and industry.The logical architecture of the knowledge graph is divided into pattern layer and data layer.The data layer describes various types of entities and relationships in the form of graph structure,which contains abundant knowledge.The analysis of the structure of the data layer map of the knowledge graph and the mining of the basic patterns in the complex structure features of the data layer will help researchers to understand the complex connection mode of the ontology in the pattern layer of the knowledge graph,and play an important role in the analysis and understanding of the knowledge graph.Aiming at the heterogeneous and sparse characteristics of the data layer of the knowledge graph,this thesis studied the sparse representation method of the knowledge graph based on Boolean matrix decomposition on the basis of constructing the hierarchical Boolean matrix model,realized the accurate extraction of the atomic structure of the knowledge graph,and improved the sparse representation efficiency of the knowledge graph.The main contributions of this thesis are as follows:1.A sparse representation method for complex networks based on Boolean matrix decomposition is proposed.Due to the floating point number generated by matrix decomposition,the existing sparse representation methods of complex networks cannot accurately extract the real atomic structure in the original network.To solve this problem,this thesis proposes a sparse decomposition algorithm based on Boolean matrix,which makes the dictionary matrix and sparse coding matrix obtained by sparse decomposition consistent with the Boolean type in the original network adjacency matrix.Experimental results show that the proposed method can accurately extract the real atomic structure in the network compared with the classical sparse representation method of complex network while ensuring the representation accuracy,and provides effective theoretical support for the sparse representation of knowledge graph.2.A sparse representation method of knowledge graph based on hierarchical Boolean matrix decomposition is proposed.Aiming at the heterogeneous characteristics of the data layer of the knowledge graph,hierarchical Boolean matrix was used to model the hierarchical Boolean matrix.Combined with the sparse characteristics of the data layer of the knowledge graph,the hierarchical Boolean matrix was divided into several homogeneous and single-layer sub-matrices,and each sub-matrix was sparsely represented by the complex network based on Boolean matrix decomposition.Furthermore,the atom fusion algorithm was used to fuse the homogeneous atoms sparsely decomposed to obtain the final atomic map.Experiments show that the sparse representation method of knowledge graph based on hierarchical Boolean matrix decomposition can effectively and accurately represent the knowledge graph sparsely,and capture the basic heterogeneous structure patterns in the knowledge graph.3.An efficient sparse representation method of knowledge graph under star constraint is proposed.Sparse representation methods of knowledge maps produce a large number of redundant atomic map structures with low usage frequency,which is not conducive to association analysis and understanding of knowledge maps.In order to solve this problem,this article is based on knowledge atlas star structure characteristics,analysis of the structure of star in the important position of knowledge map data layer,on the basis of knowledge map for star structure modeling,numerical split from the matrix,using Boolean matrix decomposition algorithm for hierarchical decomposition,will eventually atomic fusion,the efficient representation of knowledge graph based on star structure is realized.Experimental results show that,compared with sparse representation method of knowledge graph,sparse representation method of knowledge graph under star constraint can efficiently represent main structures in knowledge graph.