Research On Sentiment Analysis Of Review Texts In The Automotive Field

With the rapid development of modern science and technology Internet promoted the exchange of information.More and more consumers like to share and communicate their personal evaluation and preference of products through the Internet,which has produced a large number of comment texts.Most of these texts appear in e-commerce platforms,food delivery platforms,and some community forums,such as the common Taobao,Meituan Takeaway,and Auto Home.These text reviews often contain the emotional tendencies of the reviewers on the product.Therefore,emotional analysis of product reviews can not only provide consumers with guiding suggestions,but also help companies to obtain consumer preferences in a timely manner,and then guide business decisions.At present,sentiment analysis tasks have made great progress in the fields of online shopping,social media,etc.,but there are relatively few related researches in the automotive field.This article focuses on online review texts in the automotive field,focusing on the challenges of sentiment analysis of online automotive review texts and the problems existing in the current sentiment analysis technology in the automotive review field.In response to this problem,this paper mainly carried out the research on sentiment analysis of automotive review texts with the introduction of the attention mechanism of the long short-term memory network model,and mainly carried out the following work:(1)Aiming at the problem of the traditional machine learning method's weak ability to fit the features of big data text,and the problem of convolutional neural network's own difficulty in obtaining the sequence relationship of text data when processing sequence data.In the research on sentiment analysis of online comment texts,this paper proposes a model based on bidirectional long short-term memory networks and introduces attention mechanism into it to realize the integration of local feature extraction capabilities into serialized feature information extraction.In addition,this article explores the influence of adding emotional information in the word vector training stage in the model design work,so that the model can improve the effect of text sentiment analysis tasks;(2)Use web crawler technology to crawl online reviews related to the automotive field,and perform data cleaning and labeling on the obtained review text to achieve the construction of an online review data set for the automotive field;in addition,through the analysis of the crawled web text,it is found that the existing emotional dictionary is not enough to cover the new emotional words that are constantly emerging on the Internet.In order to effectively improve the accuracy and applicability of sentiment analysis tasks,this paper proposes a method for constructing a sentiment dictionary,this method uses these emotional words that appear in the review text but are not included in the basic emotional dictionary to build a new domain emotional dictionary,thereby enhancing the coverage of the basic emotional dictionary;(3)Carry out training experiments on the collated and constructed automotive review text data set in the above model,so that the model can better extract the features of the automotive review text,and set it comparing experiments,optimize and improve the accuracy of the model in the sentiment analysis of text reviews in the automotive field.The method proposed in this paper can finally achieve more than 90% in the classification accuracy of text reviews in the automotive field,which has certain application value.