| The popularity of the Internet and the growing scale of Internet users have supported the huge development of E-commerce in China,and online shopping is favored by consumers.With the development of social media,e-commerce has gradually formed a new development model of social commerce,and users’ online comments have also promoted the development of e-commerce.Abundant information is contained in the massive comment text,which provides help for consumers to make decisions,merchants to improve products,select marketing strategies and upgrade user experience.Fully mining online comment information is helpful to generate huge economic value,and also provides decision-making support and management inspiration for corporate brand managers.Faced with massive online comment text,it is a large task and inefficient to process and analyze it solely by human resources.Therefore,it has become the focus of e-commerce website research to crawl,process and experiment comment data and extract valuable information through text mining technology.In the field of natural language processing,the application of machine learning methods to emotion analysis and text classification is an important research topic,which still needs to be further explored,and there is room for improvement of automatic text classification algorithms.This paper proposes a sentiment analysis method based on product characteristics from sentiment analysis and text classification,and builds an improved online review classification model based on sentiment analysis to analyze online review texts.In sentiment analysis method,Word2 vec model is used to transform the text into quantization,and combined with manual extraction and Word2 vec,product feature lexicon and sentiment lexicon are established to complete the identification of ‘character-viewpoint’ pair,mark emotion polarity,and calculate emotion score.Based on this,this paper combines grey correlation analysis with Naive Bayes algorithm to consider user sentiment orientation in online comments.In order to improve the classification performance of the traditional Naive Bayes classification model,the grey relational analysis method is used to calculate the emotional correlation degree of the review text,and the grey relational analysis result is embedded into the Naive Bayes classification model as a feature attribute.In order to test the validity of the proposed sentiment analysis method and text categorization model,a real online comment mining is conducted on the Dyson V10 Fluffy Extra handheld wireless vacuum cleaner of JD.com.From the perspective of product features,this paper studies users’ emotional tendency of positive and negative online review texts,and compares the text classification results of the model with those of the traditional naive Bayes text classification model.The results show that the accuracy rate,recall rate and other evaluation indicators have been significantly improved.The model proposed in this paper has significant advantages and can be used in the practical application of online reviews,providing suggestions and management enlightenment for merchants and platform managers. |
