Research On Factual Intelligent Question Answering Method Based On Knowledge Graph

Intelligent Q&A is an advanced form of information service,which means that the computer automatically answers the questions raised by users through the analysis of user questions.Knowledge graph is essentially a semantic relation network around specific entities,which is composed of various entities,concepts and their relationships.Injecting the rich and huge real-world knowledge of knowledge graph into the intelligent Q&A model will further improve the representation ability of the intelligent Q&A model,and then,when facing the factual Q&A tasks,the recognition of user intention and entity in question answering will be more in line with the common sense of the real world,so as to improve the performance of the intelligent Q&A system.Firstly,this paper introduces the construction of domain knowledge graph and the technology status and development of Intelligent Q&A technology based on knowledge graph at home and abroad.For the follow-up research,this paper constructs a knowledge graph of automobile field,introduces the process of data acquisition and the method of knowledge graph construction.Based on the constructed knowledge graph,this paper studies the question answering method based on the knowledge graph of automobile field,proposes and implements a graph matching method,a template matching method and an information retrieval method.The graph matching method first parses the user’s question,constructs the dependency tree,and then transforms the tree structure into the corresponding semantic graph.The semantic graph is composed of nodes and relationship edges,which can be regarded as the subgraph of knowledge graph,and then matches the semantic graph with the whole knowledge graph to get the question answer and complete the answer;Template matching method needs to define the question template in advance.When the user enters the question,according to the identified trigger words and key information,the template corresponding to the user’s question is found.The filled question template is obtained by filling the slot,and then the graph database query statement is generated,which is executed in the knowledge graph to get the answer of the question;The information retrieval method first establishes an index in the text corpus.When the user enters a question,the user’s question is segmented to get the key words in the question.According to the retrieval results of the key words in the index file,the answer with the highest matching degree is returned.Experiments show that the method proposed in this paper has a good effect on answering factual questions.This paper also studies the method of generating answers only through the user input when the answer to the user’s question can not be found through the knowledge graph or the user input is not a question.Specifically,an end-to-end model based on deep learning is trained as a supplement to the knowledge graph question answering method.When the user’s intention is to chat or cannot find the answer to the question by using the above method,the end-to-end model is used to generate answers based on the user’s question content.In order to determine the intention of user input,a text classifier based on Fast Text model is trained in this paper.The user input is classified into two categories: chatting or asking questions in automobile field.According to the classification results,the corresponding question answering method is called.Finally,this paper designs an intelligent question answering system based on automobile domain knowledge graph,which integrates the previous research results and realizes two kinds of visual interaction interfaces.