Research And Implementation Of Question Answering System Based On BERT And Syntax Rules

Government services involve all aspects of people ’ s lives,and the approval items contained in the corresponding government websites are also relatively confusing.Even in the fields of employment,housing,education,and medical care that are relatively concerned,it is difficult for the people to obtain sufficient information and effective and convenient information.Solution information.If we can build a quick question and answer system based on big data in government services through the current strong information technology capabilities,then government departments can follow the trend of simplification and efficiency,and advance with the times to achieve "online management" and "online management".The high-efficiency handling mode of "one-time handling" realizes the purpose of mass users spending less time and running errands,and quickly understand the government’s government service handling policies,handling materials and handling procedures in the corresponding fields.In this paper,combined with the data of government service related office files,research and construction of the government service automatic question and answer system.The main work contents are:1.Construct a corpus of models required for the experiment,determine the language model used,design the overall software architecture,and divide the modules to simplify the development process.Propose a short text classification method model(Con-Att-Bi LSTMs)based on the self-attention mechanism to study short text classification,and finally minimize the loss rate through the softmax function.The purpose of this experiment is to provide an excellent classification model for intelligent question answering,especially in the use and research of the Bi LSTM model,to explore the feasibility of using better classification methods.2.Propose an automatic question answering method for government services based on BERT and syntax rules.Firstly,collect government affairs-related questions and answers to establish a knowledge base for government service Q&A;secondly,use the question similarity calculation method based on BERT and syntactic rules to calculate similarity;finally,ask for answers to the question and answer system based on the results of similarity calculations.The accuracy of the question and answer is evaluated and verified.3.Build a knowledge base.The data in the knowledge base mainly comes from the work items and approval items publicized on the provincial government service network.Then all the data is summarized and stored,and stored in the form of triples of inquiry content,type,and answer to the question,forming 35 departments.4.Using the content and methods studied in this article,design a question answering system design scheme based on BERT and syntax rules,and implement the system.