Sentiment Analysis Of Social E-commerce User Based On Cross-media

With the development of e-commerce and the widespread use of social platforms,social e-commerce based on social networking platforms has become a new trend for shopping.The usage of social platforms has increased year by year,while the number of potential users of social e-commerce has also increased,leading to a large amount of product review data that provide users with many reference values.However,it is difficult for people to find the desired information from a huge amount of data and intuitively analyze the pros and cons of the product.Therefore,analyzing and excavating the sentiment tendency of commodity review information in social e-commerce is of great significance to the research of commodity reputation and product recommendation.Through a large number of related literature in the field of sentiment analysis,we can conclude that mainstream e-commerce commentary sentiment analysis technology is text-based sentiment analysis.However,product reviews of social e-commerce are different from that of traditional e-commerce,Most of their review data types presented in the form of a combination of text and emoticon.The sentiment analysis method based on text information is not sufficient to obtain the emotional polarity of the comment.Thus,due to the multimodal characteristics of social e-commerce review data,cross-media data fusion methods are used to analyze the emotional polarity of social e-commerce comments.Since the emotion features of emoticons can make up for the sparseness of text,the accuracy rate of sentiment analysis is improving.The cross-media data fusion for social e-commerce commentaries is mainly based on decision-level fusion and feature-level fusion.Based on previous research,this paper introduces a method based on Canonical Correlation Analysis(CCA)to fuse the extracted texts,and eigenvectors of emoticons.At the same time,a cerebellar model articulation controller(CMAC)based on Gaussian function is served as a classifier of fusion features.The feature vector of CCA model is used as a training sample of CMAC classifier to adjust model parameters and construct information applicable to text and emoticons classifier.Compared with the traditional single-text sentiment analysis model,this model greatly improves the accuracy of social e-commerce commentary sentiment analysis.The CCA-CMAC model used in this paper have a higher accuracy in comparison to other fusion models.Experimental results show that the proposed method has good performance in both accuracy and recall of sentiment classification.