Research And Implementation Of Knowledge Graph Reasoning Technology For Public Security Intelligence Analysis

Knowledge graph is a structured database proposed by Google.It stores the relationship between entities in the form of triples.These entities and relationships are represented in the form of vectors in the database.Its purpose is to describe the semantic relationship between entities with a simple structured information,so as to achieve the purpose of logical reasoning.However,with more and more sophisticated knowledge extraction technology appeared in recent years,the amount of data contained in the knowledge graph is also increasing,and the semantic information between the structure and the entity is more complicated.Therefore,the knowledge reasoning technique for reasoning on the knowledge graph is A big challenge.How to apply the structural information of knowledge graph to knowledge reasoning and how to apply knowledgegraph and knowledge reasoning to other applications(such as question and answer systems)has become a very important challenge.Starting from the challenges faced by the current stage of knowledge reasoning,this paper studies a new knowledge representation technology(TRNG)that contains map structure information through the ripple network and applies it to the question and answer system.In addition,the construction of knowledge graph in the field of public security has been completed for the intricacies of data in the field of public security is not conducive to machine learning.It includes the following three parts:1.By referring to the diffusion process of the ripple network,and applying it to the structural information of the knowledge map.In this paper,not only the custom diffusion function is given,but also the objective function and loss function which can contain the information of the entity structure of the knowledge graph.Finally,the availability and superiority of TRNG knowledge reasoning technology are proved by experiments.2.A natural language question answering system is introduced into the knowledge graph for the singularity problem of the current knowledge graph reasoning input and output.By adding two attention mechanisms based on the knowledge subgraph at the input and output of the codec model,on the one hand,the model’s understanding of the semantics in the question is improved,and on the other hand,the quality of the answer at the output is enhanced.Finally,by training on open experimental data and comparing with other models,the natural language question answering model(QAG)based on the knowledge subgraph attention mechanism proposed in this paper is excellent in semantic understanding and dialogue model.3.From data acquisition to triple extraction to knowledge fusion,this paper designs and implements the construction of knowledge graph in the public security field,and the knowledge representation of the knowledge graph is completed by using the TRNG algorithm and successfully applied to tasks such as relationship discovery.Then,the natural language question answering system is connected to the Chinese knowledge graph,and the natural language type input and output of the knowledge graph is realized.