Financial Intelligent Question Answering System Based On Knowledge Representation Learning

At present,most knowledge graphs generally suffer from data loss and incomplete knowledge,which seriously affects the accuracy of intelligent question answering systems.At the same time,real-world information is constantly changing,and the system may not be able to accurately answer new entities outside of the knowledge graph.In the financial field,due to the traditional search method being unable to quickly obtain accurate answers,users need to consult financial personnel offline.However,due to time and personnel limitations,users may not be able to provide answers in the first place.In response to these issues,this article introduces knowledge completion technology into question answering systems,proposes a completion method that integrates entity description and neighborhood information,and an open world knowledge completion method.Based on this technology,a financial intelligent question answering system based on knowledge representation is developed to achieve accurate question answering and maximize user satisfaction.The main work of this article is as follows:(1)Propose a completion method that integrates entity description and neighborhood information.By using a graph attention network to aggregate neighborhood information,capture the semantic features of each target entity’s neighborhood,and fuse them with the text description information features of the entity,the knowledge expression ability is improved,and the accuracy of knowledge completion is further improved.The experimental comparison shows that this method has a certain improvement in evaluation indicators on both datasets.(2)For entities in the open world,without neighborhood information,it is not possible to use appeal methods for completion.To solve this problem,this paper proposes an open world completion method,which uses the joint representation and Initialization vector representation of existing entities in the knowledge map to train and generate a conversion function,which is used to map external entities that cannot obtain neighborhood information or text description information to the same joint representation space as the coder.After experiments,the model was tested on an open world dataset using MRR,Hits@1,Hits@3,Hits@10.Both indicators have improved.(3)This article develops a financial intelligent question answering system based on knowledge representation,using financial knowledge graph as data support,and using the knowledge completion based question answering method designed in this article to achieve intelligent question answering in the financial field.The front-end page is designed using HTML hypertext markup language and bootstrap framework,and data interaction with the back-end is achieved through Javascript language.The backend is developed using the Java language,and the Springboot framework is used to simplify the spring development process,achieving automatic configuration and related logical interaction.Finally,by testing the functionality and performance of the system,the results show that the system can achieve precise user Q&A and maximize user satisfaction.