Research And Implementation On N-ary Relation Mining In Knowledge Graph

Knowledge graphs are typically represented as a set of binary relations between two entities or one entity and a value.However,many relations about the world involve more than two entities,these relations are called n-ary relations.The awareness and understanding of n-ary relations will be helpful to the analysis,utilization of human knowledge in a higher order.To address this issue,many researches have been carried on finding n-ary language patterns in unstructured text.However,In this paper,we observed that n-ary relations can also be identified in graph-structured data,such as knowledge graphs.At present there has not been any work to mine n-ary relations in knowledge graph.In knowledge graph,binary relations that constitute n-ary relations frequently co-occur.In this paper,we utilize the idea of frequent subgraph mining to propose a method of n-ary relation mining in the knowledge graph:NR-Miner.The main contents of this paper are as follows:1.Propose the pattern expansion method of n-ary relation.Firstly,in order to solve the problem of current FSM(frequent subgraph mining)algorism cannot run on multilabel graph data,a multi-label pattern expansion method with versatility is proposed.Secondly,this paper analyzed the difference between n-ary relation and co-occurrent binary relation,relation node selecting method and n-ary relation pattern expansion rules are proposed.2.Propose the computation method for n-ary relation support.Firstly,a multi-label pattern support: MMNI with versatility is proposed.Secondly,characteristics of n-ary relation are used to propose a n-ary relation support: NMNI;Then theoretically prove the property of MMNI and NMNI;Finally,instance enumeration optimization in the support computation process is proposed.3.Propose the n-ary relation pruning method and carry out research on closure of n-ary relation pattern.First,because current support definition cannot fit multilabel graph data,we utilize the idea of DFS code,and propose a method for isomorphic n-ary relation pattern judgment which is used for pruning between isomorphic n-ary relations;Second,the effect of other pruning methods on pruning frequent co-occurrence binary relations in the mining process.Finally,the closure analysis is performed on the frequent patterns which are achieved by frequent n-ary relation pattern mining module,and the non-closed pattern filtering method is designed to obtain frequent n-ary relations which are closed,frequent,connected and integral.Experiments on common knowledge graph show that the n-ary relation mining method:N R-M iner proposed in this paper can accurately,comprehensively and effectively mine the n-ary relation in knowledge graph,which provides a new approach for the research of multi-relationship mining.