Research On Sentiment Analysis Of Online Reviews Based On Factors Of Usefulness

In recent years,with the continuous popularization and development of Internet technology,people can publish their views and opinions on news,products,policy guidelines,etc.on the Internet anytime and anywhere.The information that people publish on major social media not only plays an important role in the decision of the user to purchase the goods,but also helps the merchant to understand the user experience and adjust the product performance,quality and service according to the user's opinions.It is of great guiding value for the government to observe public opinions after formulating policies,adjust and perfect policies in a timely manner,and promote the harmonious development of society.However,with the exponential growth of commentary information,how to extract valuable information from these subjective commentary data has become a hot topic in current research.Sentiment analysis,also known as tendency research,identifies sentiments and emotions strength through the processing of sentences,words,and phrases.The sentiment analysis field mostly studies the useful information carried by the review texts from both the dictionary-based and machine-based learning directions,obtains the emotional score of the review text through various rules,or obtains the emotional classification through machine learning algorithms.But the dimension involved in sentiment analysis is to deal with the content of the review itself,without considering which factors contained in the review will affect the attitude and decision of the purchaser,which factors are more attractive to the user's purchase decision,and what kind of comments the user thinks more persuasive,so this thesis attempts to conduct a sentiment analysis from a different perspective,enabling it to add online commentary influence factors to sentiment analysis,and to explore this issue in in multiple.The main work of this thesis are as follows:(1)Aiming at the problem of traditional affective strength calculations which does not deeply consider the influence of modifiers on the polarity words,a method of calculating the sentiment polarity based on the review text is proposed.This thesis first analyzes the part of speech and sentence structure in the comment text;then it deeply studies the combination of various modifiers and emotional words in the review data.Since the influence of different word combinations on the difference of polar words is calculated for equal weight in the traditional affective intensity method,so this thesis proposes a method to calculate different weights according to different word combinations.The experimental results show that the method of calculating the sentiment polarity proposed in this paper is better than the traditional method based on sentiment lexicon and rules.(2)For the previous sentiment analysis study,starting from the content of the review itself,without considering the impact of the usefulness of the review on the user's purchase decision,a sentiment analysis method based on the factors affecting the online review was proposed.Firstly,the theoretical basis of the usefulness of the review is introduced,and a theoretical model conforming to the research content of this paper is proposed;Then,the relevance of the proposed theoretical model is verified using correlation analysis,regression analysis,etc.;Next,the representative product type and comment length are formulated in the sentiment analysis.The rules of usefulness,such as extremity,are reviewed;Finally,the emotional polarity calculation method is used to obtain the final emotional tendency score.The experimental results show that this method can help to correct the discrimination of emotional tendencies and highlights the analysis and mining of the value contained in the review.