Research On Sentiment Analysis For E-commerce Review Text

With the profound impact of e-commerce on people's lives,the research on sentiment analysis of e-commerce reviews has also become a hot research direction.The affective analysis method of comment text is mainly divided into the method based on the emotion dictionary and the method based on machine learning.However,the traditional affective analysis method has many shortcomings in the aspects of feature representation and extraction,and the rich and effective feature extraction is the key factor to improve the accuracy of the emotion discrimination.In order to solve the problems of the traditional analysis model in the construction and use of e-business text,this paper proposes two models for traditional analysis and tests the effectiveness of our improvement in real Chinese E-business reviews.The main work of this article is as follows:First,an improved analysis model based on emotional dictionary is established by using template matching to extend the emotion dictionary and improving the calculation method of emotional value according to different sentence patterns.Firstly,this paper collects the existing affective dictionaries,collects network dictionaries Using template matching to expand the emotional dictionary,then we can obtain emotional words in the field and construct an emotional dictionary for e-commerce reviews.Secondly,the decision tree is introduced to classify different sentence patterns,and the sentences are divided into four categories,including affirmative,negative,double negative and rhetorical,and the calculation method of emotion value is designed for each type of sentence pattern.The two kinds of sentences,which are double negative sentences and rhetorical questions,are weighted.The sentiment value of the whole comment is calculated to get its emotional orientation.Finally,it is proved that the effect of the improved model is obviously better than the traditional model by the experiment on the comment data of the Jingdong.It fully embodies the effectiveness of the improved model proposed in this paper.Second,an improved analysis model based on machine learning is established through vector feature of fusion words.The traditional machine learning based analysis method has the feature sparsity problem when it comments on such short texts.Word vector technology can realize the text's quantized representation,vector dimension is low,and it can effectively model the semantic relations between words.In order to extract more abundant and accurate text features,based on the first part of the emotional dictionary,we construct emotional words vector based on the vector technology of word vector technology,and use the TF-IDF value as the weight of the word vector to represent the importance of word in the text,finally the word vector is carried out.Weighted superposition to form a quantized representation of the text.On this basis,the model output based on the emotion dictionary in the first part is used as the feature,and the emotion analysis is carried out by the machine learning method.The experimental results show that the improved machine learning based sentiment analysis model has a further improvement in performance compared with other basic models.