Design And Implementation Of After-sales Question And Answer System Based On Machine Reading Comprehension

With the increasingly fierce marketing competition,after service becomes more and more important for the marketing of various enterprises.Aiming at the problem that traditional after service technology not only consumes a lot of manpower and material resources,but also the service effect is not very satisfactory.This paper proposes a method based on machine reading comprehension combined with attention mechanism to realize automatic interactive after service.This method mainly learns product manuals by machine autonomously,and can automatically answer various questions about product use and maintenance raised by users.This will provide a new solution for product after service.The main tasks completed in this article are as follows:(1)For user input problems and free text libraries composed of product manuals,Chinese word segmentation is processed.The word2 vec tool is used for vector representation of words to construct text features.The text features are trained through a bidirectional LSTM network combined with a reading mechanism of attention mechanism.Then it is determined whether the question input by the user matches the candidate answer,and the candidate answer with the highest matching degree is returned to the user as the correct answer.(2)The question and answer system designed and implemented in this paper mainly include four modules: Web service module,database table structure design module,question preprocessing module,and reading comprehension module.Among them,the Web service module is composed of Web pages and back-end service modules,which are mainly responsible for receiving user Q & A requests and displaying Q & A results;the database table structure design module is mainly used to store user information,historical search questions,expand the corpus;The problem preprocessing module is implemented through the interface of existing tools;the reading comprehension module uses a bidirectional LSTM combined with a reading mechanism of attention mechanism to train the corpus to obtain the final answer to the question.(3)Finally,the function and performance of the system were tested.The functional test showed that the corresponding functions of each module of the system were in line with the expected effect.The performance test mainly tested the precision of the system's answer,the recall rate and the F1 value.The experimental results are ideal.The reading comprehension model adopted in this paper performs well in the question answering system,and it also proves that the system has certain practicality.