Emotional Tendency Analysis Of User Online Reviews

With the vigorous development of the Internet,the behavior patterns of consumers have changed deeply.The researches and analyses of online product reviews,in particular,using sentiment analysis to judge the attitudes of the reviews,are important to consumers and businesses.Online product reviews include texts and grades.Grading can clearly and simply indicate the attitudes of the purchased users,but it can only indicate the consumer's emotional orientation roughly,and different users have different grading standards.To understand the distribution of sentiments in online product reviews accurately,it is important to use text for sentiment analysis.The purpose of this paper is to use sentiment analysis techniques to analyze the attitudes of the reviews on the mainstream auto websites in China,and obtain the scores of sentiment polarity degrees,and combine with the grades to calculate the review scores synthetically.The main innovations of this paper are as follows:1.We optimize the traditional method of emotion classification based on sentiment lexicon and calculate the sentiment polarity scores of auto online reviews.(1)We analyze the grammatical relations of words by means of Dependency Parsing.And use the rule matching method to extract the triplets of commodity attribute words,sentiment evaluation words and sentiment degree words in the auto online reviews simply and effectively,(2)We extract the sentiment words from the HowNet sentiment lexicon and the sentiment words labeled manually as the sentiment seed words.And then we use Word2Vec to calculate the semantic similarity,between the sentiment seed words and the candidate words.Based on the semantic similarities,we get the sentiment polarities of the candidate words.Finally,we give the sentiment weights of the sentiment words to create sentiment lexicon.2.We propose SCCDW(Sentiment Classification by Combining Dependency Parsing and Word2Vec)based on machine learning,which classifies the sentiment polarity of the auto online reviews,to modify the sentiment weights of sentiment words.(1)We extract the word pairs with grammatical relations as features based on the Dependency Parsing,which avoids using the same vector to represent sentences with the same words but the different grammatical structures.(2)In order to reduce the dimensionality of the feature vectors and keep the contributions of the low-frequency words,we obtain the distributed vectors learned by Word2Vec and group the semantic similar words in a cluster through the K-means to obtain the pairs of each word and its corresponding cluster,and then replace every word with its corresponding cluster label.3.We propose a comprehensive evaluation method of auto online reviews based on the sentiments of the review texts.Combined with the sentiment polarity calculated by SCCDW method and sentiment polarity degree scores calculated based on the sentiment lexicon,to modify the sentiment weights of sentiment words.Finally,we calculate the sentiment polarity degree scores of the review texts and divide the scores into grades,to calculate the comprehensive score of the auto online reviews.In this paper,we use auto online reviews on the Auto-home website to validate the method presented in this paper through experiments.The results of experiments show that the method we proposed can effectively analyze the sentiment polarity of auto online review texts and calculate more reasonable comment scores.