Research And Implementation Of Tax Consultation Question Answering System

Tax consultation has a large volume of business and requires a high degree of professionalism.In recent years,there are many new tax preferential policies,and the demands for tax consultation grow fast.The existing ways of tax consultation are mostly telephone consultation and on-site consultation in the Tax Service Hall.Tax workers need to study relevant policies constantly,and reply to a large amount of consultation,which makes the tax authorities face great pressure.With the rapid development of computer application technology,some tax authorities have begun to try to use the Internet for on-line tax consultation.However,most of them are message consultation,which requires manual answers to the questions,so a large number of hot issues may cause system congestion,and the answers are not timely enough to help people solve the current problems effectively.As a cutting-edge technology,the intelligent question answering system has been applied in many fields and achieves good results.By searching and learning the existing intelligent question answering algorithm model,we research the key technologies that could be applied in the field of tax consultation answering,discover some problems that these question answering technologies may have in the field of tax consultation,then find the solutions and design a Tax Advisory Question Answering System.Based on the key technologies of intelligent question answering and combined with the popular deep learning algorithm,the system can answer the questions related to tax policies immediately,improve the efficiency of tax consultation and reduce the pressure of tax authorities.Finally,this thesis implements a question answering system for tax consultation,and research the core technology of the system.The system has the advantages of high efficiency and strong practicability,which provides more convenience for taxpayers and tax workers.We mainly use two technologies: Information Extraction and Machine Training,and take system management and interaction as the core part.The system use Jieba word segmentation technology and the Word2 Vec tool to extract information,CNN(Convolutional Neural Network)algorithm to train the machine,Regularization,Re LU activation function and learning rate annealing are used in the training process.Besides,in order to improve the development efficiency,we use Micro Service Framework for development,Spring Boot Framework for logic development,Spring Cloud Framework for service management,and Spring Security as a Single Sign-on security framework.The system program structure is clear,the development process is convenient and fast,the use of the above technology can improve the stability,operability and safety of the system.