Design And Implementation Of Financial Intelligence Question Answering System Based On Knowledge Graph

The research of QA system is one of the important attempts to cross the semantic gap between man and machine.Human language is fuzzy,diverse and imprecise,while computer takes a clear,structured form of expression.Using knowledge graph to store massive factual knowledge provides information basis for QA based on knowledge graph.At present,with the rapid development of QA based on open domain knowledge graph,there are many mature solutions.However,few researchers try to introduce knowledge graph into financial field,construct question answering system and carry out daily financial consultation.The design and implementation of financial QA system based on knowledge graph is build a knowledge graph of financial domain,and design and implement the QA system on this basis.The system grab the data of financial website,and integrates the existing financial data within the enterprise.Through the analysis of these data,it constructs the financial domain dictionary,and determines the entity and relationship definition of the knowledge graph in this paper,and constructs the financial domain knowledge graph.Then,the multi pattern matching AC algorithm is used to realize entity recognition and question template based intention recognition to realize QA Based on knowledge graph,and training word2 vec network and calculating text similarity is used to realize the QA function of retrieval based on question answering knowledge base.By comparing the matching efficiency of multi pattern matching AC algorithm with other multi pattern matching algorithms,the effect of entity recognition based on dictionary multi pattern matching is verified.It is found that reversing all pattern strings and constructing finite automata to carry out multi pattern matching from back to front takes more memory,but it takes less time.Moreover,the intention recognition based on question template has higher recognition rate for simple questions.For economic issues,the accuracy rate of answer is as high as 85%.And because of the implementation of the question answering knowledge base retrieval function,the system can give the answers to similar questions for complex problems,and the overall effect is satisfactory.At present,there are 2077 entities and 2072 relationships in the financial domain knowledge graph,which has covered most of the enterprise financial Q & A knowledge.For the problems that cannot be solved by the system,the system provides a way to import knowledge manually.The administrator can update the knowledge graph of the financial field through the operation chart database.The system can use the new knowledge to answer the new problems,and finally form a closed-loop finance Intelligent QA system.