Chinese Sentiment Analysis Based On Feature Segmentation

As a necessity for people's lives,mobile phones has a wide variety of brands and styles.Because of the information asymmetry between buyer and sellers,it is a agonizing thing for users to choose one brand on network,and there needs to be more specific findings to help them make their choices.With the development of natural language processing technology,text analysis has become a hot area,of which emotion analysis is the forefront.Textual analysis is relatively mature in the field of English studies at present.However,in the field of Chinese studies,the analysis of the Chinese texts has not reached a certain level due to the complexity of the Chinese language.First of all,the research volume of the text sentiment analysis at the sentences and chapters levels is far from enough at present,so it is very necessary to make the study go on.Second,the task of sentiment classification for large-scale datasets also needs more excellent sentiment classification model of machine learning to play its role,which we must continue to explore.Finally,the related technologies about text sentiment analysis is now more focused on weibo,news,film criticism,hotel reviews and so on,so there is an urgent need to expand the research field and dig out more research conclusions with practical value and social benefits.The article collected some comments about five popular mobile phone from jingdong Mall and taobao,then do the doc2 vec vectorization about these datasets,and use the gradient promotion decision tree algorithm to classify the emotions,find that the classification accuracy approaches 81.79% in the article.Then,by comparing this classification model with the two methods of logistic regression and random forest,I found that the AUC of GBDT is about 0.11 higher than LR and about 0.03 higher than RF,and the FPR of it is lower than that of LR and RF Nearly 20 percent,and at the same time,the GBDT have the shorter classification times than the other two models,thus verifying the superior performance of the doc2 vec + GBDT classification model.Finally,through the parameter adjustment,the accuracy of the doc2 vec + GBDT classification model is improved by 1.5% and the AUC value is also improved by 0.03.The article then use this model to make the emotional classification of 5popular mobile phone reviews,and make further analysis about the user's attribute preferences of different brand mobile phones.The conclusions of this study not only can help mobile phone manufacturers to understand customer needs and improve their own products,s,but it can also help customer to select their favorite products based on their needs.