Research On Fine-Grained Sentiment Analysis Method Of Online Shopping Comments

In recent years,with the rapid development of e-commerce,online shopping reviews,as a digital version of traditional word-of-mouth,have become one of the most important sources of information about the evaluation of network consumption.On the one hand,a large number of theoretical studies and practical applications show that online shopping review informations contain the emotional tendency of consumers to purchase goods,it not only has an important impact on subsequent potential consumers'purchase decisions,but also has a very important guiding significance for network providers to continuously optimize business strategies,improve services and merchandise quality.On the other hand,different consumers and merchants often pay different attention to product characteristics,and then lead to the result of the coarse-grained analysis based on the overall sentiment polarity analysis for online shopping reviews has been unable to meet the increasingly diverse needs of customers.Based on the above background,this paper is based on the comment data about mobile phone on e-commerce website.Emotional analysis will be acted on different attributes of products,and conducted online shopping comment information fine-grained sentiment analysis methods.Based on the summaries of emotional analysis'theories at home and abroad and their main research results,the main research work of this paper mainly includes the following three aspects:(1)There are a large number of non-opinion sentences in the original online shopping comment information.To eliminate the negative influence of non-opinion sentences on the emotion analysis,this paper proposes a method about online shopping comment opinion sentence extraction based on sentence vector and SVM algorithm.In this method,at first,using word vector tool(word2vec)which is developed by Google based on deep learning to generate the word vectors corresponding to the words in the online shopping review.And then through analyzing the contribution rate of words to sentence semantics,using word vectors and normal distribution formula to build sentence vectors.Finally,the sentence vectorsare input as the feature of SVM algorithm for extracting opinion sentences.(2)Based on the experimental corpus,on the basis of the empirical research which findclassification rules of semantic similarity about emotional words and attributes based on Euclidean distance and cosine distance,an affective word-attribute word extraction method based on semantic similarity calculation is proposed.This method treats the extraction of emotional words and attribute words as a kind of two classification problem,the semantic similarity threshold of emotional words and attribute words is used as the classification standard,the automatic extraction and matching of emotional words and attribute words is realized by combining the classification algorithm,at the same time,a corresponding extraction method is proposed pointing at the implicit attributes in the corpus.(3)Expand the emotional tendency from the traditionalcategories which include positive and negative to the four categories which include compliment,derogatory,neutral and both appraise.By improving the calculation method of the traditional word-based tendency of emotional words,the fine-grained emotional inclination is calculated.