Research On Question Answering Of Knowledge Base Based On Query Graph

With the rapid development of information and intelligence,the network search engines and retrieval methods are also changing,and a highly automated question answering system is put forward.The question and answer system can directly understand the user’s question,return a simple and correct answer,and reduce the user’s query cost.As an effective entity-relation-entity expression,knowledge graph can construct a real and objective observation graph structure for entity semantic relations in the vocabulary,with strong link characteristics and complete structural attributes.The question answering system based on knowledge graph combines the independent advantages of the two,and has become a research hotspot in the field of artificial intelligence,which has attracted extensive attention from scholars.In the process of continuous development and improvement,the Chinese knowledge graph is characterized by unique word cases such as multi-word synonym,polysemy,contextual reference,and extended meaning,so it is not good in candidate entity recall rate and candidate entity disambiguation accuracy.Based on this,this thesis designed a dictionary tree and model recognition method to improve the candidate entity recall rate,and extracted the candidate entity features through the fusion method of numerical statistics and semantic analysis to reduce the incidence of ambiguity.At the same time,because the Chinese knowledge graph relationship is complex and there are many application formats,the prediction relationship is divided in this thesis,and the improved BERT prediction model is used to discriminate entity triples in the basic relationship.In the advanced relation,the query graph is used to query the task path,and the query graph is classified into extended type and connected type.According to the node degree,three query graph decomposition methods of core,unified and branch are designed.On this basis,this thesis designs a query algorithm based on query subgraph join and matching to improve the accuracy of query target answers.Experimental tests verify the effectiveness of the algorithm.For system development,this thesis proposes a graph convolutional neural network composite model,establishes a complete application model of question answering system based on knowledge graph,and uses Baidu knowledge graph to build an open domain Chinese question answering system,which improves the overall design process of the system and realizes It has the function of displaying the query content in the form of a web page.