Design And Implementation Of Several Key Technologies Of Automatic Question Answer System

With the rapid development of artificial intelligence and natural language processing technology,question answering system has been applied in various fields.In order to make the question and answer system better serve all fields,this thesis takes the design and implementation of automatic question answering system as the subject,focuses on how to set up and implement automatic question answering system.The combination of machine learning and traditional natural language processing can improve the accuracy and automation degree of the automatic question answering system while reducing system complexity and making the automatic question answering system more robust.The system in this thesis is a retrieval Chinese question answering system based on deep learning,which is mainly divided into three parts.1?The first part is the design and implementation of the problem generation subsystem.The automatic question generation subsystem is responsible for generating the question answer to question corpus from the original document for the retrieval of the subsequent automatic question answering system.In this thesis,problem generation is divided into simple problem generation and complex problem generation.Firstly,a set of document pre-processing process is proposed to extract sentence candidate sets from documents for generating problems,a template-based and rule-based method is proposed for simple problems,and an improved problem generation model is proposed for complex problems.And then a method of scoring for problems is proposed to filter unqualified generation problems.Finally,how to use these methods to realize the problem automatic generation subsystem is introduced.2?The second part is the design and implementation of question retrieval subsystem..The question-answering retrieval subsystem is responsible for retrieving the matching question answers from the corpus based on the question answers generated by the question automatic generation subsystem and the questions input by users.Firstly,in order to speed up the retrieval speed and accuracy,the search engine was introduced,and the candidate standard problem set was obtained by using the scoring mechanism of the search engine,and the candidate standard problem search score was obtained by using the scoring mechanism of the search engine.Then an improved question retrieval model is proposed to calculate the semantic similarity score of user input question and question answering corpus.Finally,how to use these methods to realize the question retrieval subsystem is introduced.3?The last part introduces how to integrate the two subsystems into an automatic question answering system.In order to improve the scalability and maintainability of the automatic question answering system,the automatic question answering system is further decoupled.The automatic question answering is divided into three functional modules: preprocessing module,automatic generation module and question answering retrieval module,The implementation of each module and how to interact with each other are introduced.On the basis of the above methods,this thesis designs and implements an automatic question and answer system,and conducts a comparative test on the system.The test results prove that the automatic question and answer system designed and implemented in this thesis is effective..