Intelligent Q&A Based On Domain Knowledge Graph

The topic of this thesis comes from the national key research and development program"construction and demonstration application of public cultural resources intelligent construction and sharing and management platform(2019YFC1521405)".In the field of Knowledge Q&A,people begin to seek more intelligent way to obtain knowledge.At present,Intelligent Q&ABased on knowledge graph uses single-layer network for intention recognition,and theability to recognize deep semantics needs to be further strengthened.At the same time,withthe increase of user data and the emergence of new problems,more and more types ofproblems need to be solved,and the intelligent Q&A system needs to be expanded with thechange of time.Therefore,on the one hand,how to better identify the deep semantics ofusers,on the other hand,how to make the intelligent Q&A system dynamically expand thequestion type,has become two urgent problems in the field of intelligent Q&A system.To solve the above problems,based on the research of user intention recognition and dynamic expansion,this thesis studies the double-layer user intention recognition technology and adaptive expansion method based on keyword separation,and realizes the deep-level user intention recognition and dynamic adaptive expansion function of intelligent Q&A system.Finally,taking the subject curriculum field as an example,the system is designed and implemented to verify the feasibility and effectiveness of the proposed method.The main research work of this thesis includes:1.This thesis designed the construction process and system architecture of adaptive and extensible intelligent Q&A assistant based on knowledge graph.The intelligent Q&A assistant based on knowledge graph is divided into four modules,namely "question processing module","answer processing module","adaptive expansion module" and "data processing module",which can be extended adaptively on the premise of meeting the basic needs.2.This thesis designed and builded the case base of user typical questions as the retrieval rule base of intelligent Q&A assistant.The relationship between knowledge graph and user question types is explored.The course syllabus is used as the data source,the ontology model is designed,and neo4j is used as the data storage medium to build the knowledge graph.3.A two-layer user intention recognition model based on keyword separation is proposed to solve the problem that deep user intention can not be recognized by using single-layer network in the process of user intention recognition.This paper analyzes the importance of keyword features in user sentences,and then carries out two levels of user intention recognition based on this.The first level of user intention recognition obtains keywords and question sentences;The second level of intention recognition is the problem type.Combining these two levels of user intention,we can get all the user's intentions.4.The method of Intelligent Q&A assistant extension is proposed,which realizes the requirement that the intelligent Q&A assistant can still be extended adaptively when the user feedback data is insufficient at the initial stage of Intelligent Q&A assistant construction.In view of the fact that syntactic structure can effectively recognize sentence features,the syntactic structure with separated keywords is used as the substitution of user sentences,and hierarchical clustering algorithm is used to cluster user sentences automatically.Through this algorithm,different thresholds are used to get the best threshold of hierarchical clustering algorithm.5.On the basis of the above research,with pycharm as the development tool,neo4j as the data storage medium,and python as the development language,the Intelligent Q&A assistant for subject courses is designed and constructed.Finally,the system is tested to verify the correctness and effectiveness of the research method in this thesis.There are 25 figures,14 tables and 64 references in this thesis.