Research On Financial Knowledge Graph Construction And Question Answering Technology

With the development of the market economy and the information age,the people's demand for market knowledge collecting are increasing.However,the development of the knowledge graph and question answering technology for financial domain is facing a lot of difficulties,such as the variety of information sources,the complex data structure,the difficulty of text data representation and the last but not least,the difficulty of intent detection and knowledge retrieval.In order to solve such problems,this paper designed and implemented an effective method for constructing financial knowledge graph and studied key issues in question answering technology based on natural language processing and data mining technology.The main research contents are as follows:we crawled the existing semi-structured encyclopedia content,and constructed a financial knowledge graph by text classification,dependency syntax analysis and association rules.With the knowledge graph,we designed a specific pre-training task to construct the text representation for financial domain.Based on the knowledge graph and the text representation,we built a joint intent detection and entity identification model to analyze user's query.And we constructed several knowledge retrieval technologies based on the knowledge graph to obtain the knowledge triplet.Combing the above,we built a complete question and answer system framework based on the knowledge graph.The main contribution of this paper are as follows:we designed and implemented a simple and effective method for constructing knowledge graph on specific domain with semi-structured encyclopedia content,a pre-training task for text representation based on knowledge graph,a joint model to detect intent and extract entities from user's query,and a framework of question and answer system based on the knowledge graph.The experimental results show that our methods achieved satisfactory performance on tasks like knowledge graph construction,text representation,entity extraction,and problem classification.