Research And Implementation Of Intelligent Question Answering System Based On Domain Knowledge Graph

In recent years,with the rapid development of science and technology,the Internet has subtly affected all aspects of people's lives.It's accessibility has gradually increased the flexibility of obtaining information.For example,information is obtained through a search engine,and a user searches for an answer by browsing the relevant pages returned by the search engine.However,for specific fields,such as the financial field,financial products and field information are complex and diverse and the way to search and return answers is not efficient enough.Users still need to eliminate redundant information based on domain expertise,so the question answering system came into being.Nowadays,the coverage rate of financial services in my country has reached99%.More and more people want to enter the "financial circle".Traditional consulting methods are limited by geographical and time constraints.Therefore,it is necessary to develop an intelligent question answering system in the financial field.For intelligent question answering,this thesis adopts two subtasks in user intent recognition: entity recognition and entity linking.Identifying financial entities and related mentions using the BERT-Bi LSTM-CRF model in entity recognition tasks;In the entity linking task,the BERT-CNN-DSSM model is proposed to remove redundant information.The model uses BERT for text preprocessing,low-dimensional vectorized representation of entities and mentions obtained in entity recognition tasks,and input into the CNN layer.Finally,the results are input into DSSM for semantic similarity calculation to complete the entity disambiguation work,and the superiority of the model is verified by comparison experiments with other models.In view of the lack of domain knowledge graph,this thesis constructs a knowledge graph in the financial domain and clarifies the construction process: First,analyze the collected financial data,design entity,relationship and attribute tables,and then perform information extraction and knowledge fusion to obtain the knowledge map in the financial field.Then the knowledge graph is stored in the Neo4 j graph database as the knowledge database of this thesis.Based on the above main work,this thesis designs and develops an intelligent question answering system based on knowledge graphs in the financial field.The system is based on demand analysis and is oriented to financial practitioners and ordinary users.Provides functions such as system management,data management,and Q?A.After the user types in the question,semantic analysis is performed to obtain key entities,and then the knowledge database is retrieved.Return text explanations or related knowledge graphs according to user instructions.At the same time,financial practitioners are supported to operate entities in the knowledge graph to update the knowledge graph.The system adopts B/S architecture and uses Python for development,VUE.js completes front-end page functions,CSS for simple beautification,and the database includes Mongo DB,Neo4 j and My SQL.Through functional testing and performance testing,the system in this thesis can realize intelligent question answering in the financial field,and the system meets the expected standards in both function and performance.