Research On Automobile Product Selection Method Based On Online Comment Text Analysis

In recent years,with the rapid development of the Internet,online shopping and social media have become popular rapidly.Users generate a large number of text reviews on automobile products and other aspects on the Internet.These user data have a lot of valuable information,including the user's satisfaction with the experience of the product in the process of use.Using appropriate text analysis technology to classify emotions can understand the user's demands from the comment text and provide a reference for new consumers to purchase.However,there are relatively few studies on emotional classification and product selection of automobile online reviews,especially the lack of in-depth learning and other technologies to explore and improve this issue.In this regard,aiming at the problems of emotional classification and product selection methods of online automobile reviews,this paper carries out research on emotional classification and product selection based on threshold recursive unit.It is of great theoretical and practical significance to use in-depth learning method to extract the emotional information of user experience in comment text more fully and efficiently,to help enterprises improve product performance,and to enable consumers to understand the use of products more comprehensively.The following work has been accomplished:(1)The multi-channel modeling of automobile online comment text is completed.Aiming at the characteristics of online comment text and the problems in emotional classification,word vector channel,part of speech vector channel and word emotional information channel are constructed,and multi-channel fusion is completed.(2)The F-BiGRU emotional classification model based on multi-channel modeling is proposed.In order to improve the accuracy of emotional classification of automobile online comment text,the model is improved on the basis of threshold recurrent recurrent recurrent neural network.The model extracts more sufficient semantic information from online comment text through feature enhancement layer,and uses bidirectional threshold recursive unit to extract text semantics to complete the task of text emotional classification.This model is more suitable for the randomness and colloquialism of Internet commentary text,and is compared with traditional machine learning model and convolution neural network model.The model improves the accuracy of emotional classification and can better complete the task of emotional classification.(3)The method of automobile product selection based on online comment text is put forward.Through the F-BiGRU model of multi-channel modeling,the emotional value of comment text is calculated,and the emotional value of each attribute of different vehicle types is obtained and visualized.The TOPSIS method is used to calculate the emotional value of each attribute,and the comprehensive emotional value of each candidate car is obtained,which can be used as a reference for consumers to improve their understanding of the user experience of the car and help consumers to make better choices of automobile products.