Research And Implementation Of Automatic Question Answering System In E-commerce Field Based On Deep Learning

The Internet generates a large amount of data every day.Users hope to be able to quickly obtain the required content from a large amount of data.At present,the commonly used ways of retrieving information can not fully meet the users' retrieval needs.Users want to be able to directly retrieve the content they really need,rather than retrieve a large number of related documents,and then go to check each document in turn,to find out whether the document contains the content they are looking for.Question answering system is a common form of information retrieval,which can answer questions quickly and accurately.Question Answering System(QA)has been applied to different fields.It is a hot research topic in the direction of artificial intelligence and natural language processing.Nowadays,China's ecommerce industry is booming,online shopping has become the preferred shopping mode for most users.With the increase of e-commerce users,the amount of user consultation on e-commerce websites also increases.In order to provide users with better service,we need to build a professional and huge manual customer service team.Compared with the manual customer service,the automatic question answering system can provide professional service for users 24 hours without the influence of time and environment.In the electronics industry,the use of automatic question and answer system can undertake part of the work of manual customer service,save costs for enterprises and enhance the user experience.Therefore,the research on the automatic question answering system based on the field of e-commerce has great application value.The purpose of this paper is to construct an automatic question answering system for ecommerce.The main contents of this paper include the following aspects.The core part of the automatic question answering system is the algorithm module.Aiming at the consulting characteristics of the e-commerce industry,this paper designs and implements the algorithm engine module.In this paper,an in-depth study of the automatic question answering system is carried out.Considering that e-commerce users' queries are generally goal-driven,the question answering system is finally constructed based on the form of question answering pairs.In this paper,the process of answering user's questions by question answering system is abstracted into the process of selecting the most matching questions in the knowledge base.The key to this process is the semantic matching algorithm.At present,deep learning has achieved good results in semantic matching.The training of deep learning algorithm needs a lot of data,and the final effect of the algorithm is closely related to the quality of the data set.This paper collects a large number of customer service data,and constructs a question-and-answer set in the field of e-commerce by combining manual labeling with automatic labeling.Based on the constructed corpus,this paper implements different semantic matching algorithms and compares their performance on the data set.The experimental results show that the algorithm model based on convolution neural network is the best.This paper constructs a complete question-and-answer system,and provides a complete knowledge base management module for the question-and-answer system.The question answering system implemented in this paper can be divided into three modules: question comprehension,information retrieval and answer generation.Question comprehension module classifies users' questions and analyses users' real intentions.The information retrieval module retrieves the most matching problem from the knowledge base.Answer generation module gets the answer to the question and returns the answer to the user.This paper also designs and implements the knowledge base management module for question answering system.The knowledge base management module supports adding and modifying question and answer pairs,and provides accurate training data for question and answer system.At the same time,the knowledge base management module can count the user's questions and the matching situation of the question answering system,which provides a basis for the analysis and improvement of the system.