Research On The Construction And Complement Of Knowledge Graphs In The Financial Field

With the implementation of cloud computing,big data,artificial intelligence and other emerging technologies worldwide,the amount of Internet data doubles every two years.To better analyze and apply Internet data,scholars and institutions at home and abroad have proposed the concept of knowledge graphs.The knowledge graph presents the concepts,entities and relationships in the objective world in a structured form,expresses the information of the Internet in a form that allows humans to understand the world more intuitively,and plays a huge role in the management,organization and analysis of the application of Internet information effect.At present,the knowledge map has been applied in the fields of medical,commercial and transportation industries and achieved good results.However,at present,there are relatively few knowledge graphs for the financial industry.Many companies 'financial graphs are not open to the public due to the privacy of the information and there is an implied relationship when they are applied.Therefore,after the construction of the knowledge graph is completed,the subsequent completion work is also completed.Change is particularly important.Classic knowledge graph completion models often only focus on the path formed by two adjacent nodes in the knowledge graph,ignoring the information brought by the relationship path formed by the node and other multiple nodes.In order to solve the problem of lack of financial graphs and the defects of the knowledge graph completion model,this paper first uses financial stocks and corporate information to construct financial knowledge graphs.On the one hand,information such as financial stocks is used as an example to describe the construction method of financial knowledge graphs.On the other hand Provide a basic data set for the later knowledge completion task,and then propose a knowledge graph completion method based on the combination relationship path to verify the completion task of the financial knowledge graph.The main work includes:1.Based on the current situation of the financial knowledge graph,through the bottom-up knowledge graph construction method,crawled the stock website-eastmoney.com,Baidu stock,corporate website-QCC.com?financial stocks of enterprises and relevant financial forums,legal persons Information,through natural language processing algorithms to extract and identify the relationship between business entities,legal entities and entities,and finally cleaned and imported into the neo4 j graph database,elaborated the main process of building a small financial knowledge graph,completed the construction of financial knowledge graph,and for Later knowledge completion work provides basic data sets.2.Aiming at the defect that the translation model at this stage only considers the connection between two points,based on the different 'roles' represented by the direction of the connection between the nodes,a knowledge graph completion model based on the combined relationship path is proposed to The probability of distribution describes in detail the three different connection path connection modes under the three-node structure.The vector representation of the nodes is obtained,and the model is optimized by negative sampling and stochastic gradient descent.Finally,the financial knowledge graph constructed in this paper and the Freebase dataset are used as data sources to verify the effectiveness of the method.