Research On Emotional Analysis Method Based On E-Commerce Review

With the all-round development of the Internet and the popularization of e-commerce,online shopping provides convenience to consumers,and has become a popular shopping method.The commentary information on the e-commerce website is also increasing.These commentary information expresses the subjective feelings of consumers on the purchased goods,and has great reference value for consumers to select products that are in line with their intentions.At the same time,it is also an important basis for improving the marketing strategy for merchants.With the massive accumulation of reviews on e-commerce platforms,consumers are paying more and more attention to the topic information of products they are interested in,such as the quality of products,packaging,or delivery speed.Therefore,extracting emotional information on product topics from e-commerce reviews is a hot research topic.This paper conducts sentiment analysis based on the e-commerce commentary text,mainly extracting comment feature words,including the subject words and emotional words of the product review text,and discriminating the emotional polarity values corresponding to the emotional words.There are three kinds of emotional polarity: positive,negative and neutral,which are represented by 1,-1 and 0 respectively.This article first uses the Octopus Collector software to collect e-commerce commentary texts on the Jingdong Mall website,and manually identifies the subject words,emotional words and corresponding emotional polarity in each e-commerce commentary.Then,based on the traditional sentiment dictionary method,experiments are carried out,and the deficiencies of the traditional methods are obtained based on the experimental results.Then three improved sentiment analysis methods based on the traditional sentiment dictionary are adopted,and the experiments are compared to select the methods suitable for the research.The three improved sentiment analysis methods used in this paper are as follows: First,combined with the CRF named entity recognition method,CRF named entity recognition can be used to extract the subject words and emotional words,but the subject words and the emotional words cannot be used one by one,and need to be utilized.The dictionary matches,finds the missing subject words or emotional words,and uses the sentiment dictionary to judge the emotional value,so that the subject words,the emotional word and the emotional value are one-to-one correspondence;Second,the method of combining the positive maximum matching word segmentation,firstly segment each e-commerce comment,then use the forward maximum matching method to segment each clause,and use the dictionary to match the emotional words and the subject words,and determine the emotional value.Third,based on the improved emotional dictionary method,first construct the appropriate adverb dictionary and negative word dictionary.When extracting emotional words,use the adverb dictionary to properly remove the prefix part of the non-emotional word beginning with the adverb prefix,and judge whether it is an emotional word again,and then adjust the distance between the negative words and the emotional words to judge the negative words,and the negative words and emotional words form new emotional words,and use the emotional dictionary to judge the emotional value,then adjust the suitable distance between the subject words and the emotional words,and extract the subject words with the help of the dictionary.By comparing the experimental results,the effect of the method combined with CRF named entity recognition is not obvious,however,based on the improved sentiment dictionary method and the method of combining positive maximum matching word segmentation,the effects of the two methods are significantly improved,and the based on the improved emotional dictionary method works best.