Design And Implementation Of Automatic Question Answering System Based On Knowledge Representation Learning

Today's Internet age is full of massive amounts of information,and how to quickly retrieve and obtain the required information has become a challenge.In information service applications,the question answering system is an advanced information retrieval system that can answer user questions in natural language.In order to improve the accuracy of information services,the structured knowledge graph is used as the underlying support of the question answering system to provide the system with the knowledge needed to answer questions,but this leads to the question of how knowledge is represented.Traditional question answering systems based on symbolic representation and semantic analysis face problems such as low computational efficiency and data sparseness in actual use,which are not conducive to question understanding.With the development of deep learning,knowledge representation learning technology is becoming more and more popular.Many scholars have paid attention and research,and introducing it into the question answering system to improve the accuracy and efficiency of semantic calculation.This paper starts from the application scenario of automatic question answering based on the knowledge graph.It is designed and implemented for the problems that the semantic information in the knowledge graph is difficult to be fully utilized,and the diversity and ambiguity of natural language problems raised by users.An automatic question answering system based on knowledge representation learning is proposed.Specifically,it can be divided into three research contents:First,in order to operate the knowledge graph more conveniently in downstream applications,the entities and relationships in the knowledge graph are mapped to a continuous vector space through knowledge representation learning,and at the same time contains certain semantic information;secondly,in order to achieve the correct mapping of the phrase in the question to the entity on the knowledge graph,and then generate a structured query,the mapping process is solved by combining the representation learning and the entity linking model of the deep neural network.Finally,apply the knowledge graph representation algorithm and entity linking algorithm to the question understanding module of the question answering system to realize a relatively complete automatic question answering system based on the knowledge graph.The key algorithm of this paper is to integrate the knowledge representation learning model of multi-source information and the joint entity link model based on deep learning and knowledge representation.The former integrates entity text description information and entity category information on the basis of the existing knowledge representation learning model TransE.Compared with the representation that only relies on the triple structure,it strengthens semantic relations and improves computational efficiency;the latter uses the advantages of deep learning to extract the multi-granularity representation features of mentions and candidate entities,solve the problem of entity linking and word sense disambiguation,and help the question answering system of this paper to more accurately understand the semantic information of questions.Experiments show that the two models are better than the baseline model on the tasks of knowledge representation and entity linking.In order to realize the knowledge question answering system,this article first investigates related background technologies.Then,based on user requirements and related technologies,the overall requirements of the system are analyzed,and functional requirements and non-functional requirements are listed.Aiming at the two key issues of knowledge representation and entity linking,the paper carried out research and provided solutions,and conducted experiments and analysis on the effectiveness of the model according to the evaluation criteria.Finally,on the basis of completing the research on key issues,the overall design and detailed design of the question answering system,including the interaction design of the key task workflow and the interface design of each module are carried out.After the system is implemented based on the design,the paper conducts related tests and analysis on the system,which proves the usability,ease of use and user-friendliness of the system.