Fine-grained Emotion Analysis Of Drug Reviews Based On Micro-blog

Objective:This article makes a fine-grained emotional analysis to themicroblog from the perspective of data mining,and the emotions are divided into 8 categories of “happy,like,angry,sad,fear,evil,panic,and suspiciousness” by calculating emotional intensity values,So as to restore microblog user emotions.Taking the drug-related microblog as an example,it not only explores the feasibility of fine-grained sentiment analysis,but also provides decision support for consumers to purchase drugs,which is Convenient for follow-up research.Method:Through reading a large number of documents,the main methods of the current domestic and foreign sentiment analysis and their advantages and disadvantages were analyzed,and the method of sentiment analysis combined with the rules of sentiment lexicon was determined.Therefore,this paper supplements the emotional vocabulary dictionary and constructs a series of auxiliary dictionaries to fully identify the emotional information in the microblog and strives for the accuracy of the sentiment analysis.First,on the basis of the basic emotional vocabulary dictionary,based on the“Chinese emotional vocabulary ontology library” constructed by Dalian Institute of Information Retrieval,combined with “Synonym Cilin” compiled by teacher Mei Jiaju,First,based on the “Chinese Emotional Vocabulary Ontology Database”constructed by Dalian Institute of Technology Information Retrieval,the basic sentiment vocabulary dictionary is merged with the “synonym Lin” written by Teacher Mei Jiaxuan to achieve the improvement of the emotional vocabularydictionary.;Secondly,with regard to the fact that Internet users have more emotional doubts on the Internet,a dictionary of interrogative words is formed based on the query vocabulary constructed by the Chinese Emotional Vocabulary Ontology Database,Modern Chinese Dictionary,Synonym Word Lin,and Sinamicroblog.Based on the words contained in the Chinese Emotional Vocabulary Ontology Database,the strength of interrogative words is obtained through the mutual information method.The experiment proves that the question dictionary constructed in this paper can effectively identify the emotions in the microblog text;in the same time,taking into account the diversity of sentimental expressions in social media platforms,113 emojis commonly used on microblog and 90 network term obtained after screening were selected in this paper,and compared with the “Chinese Emotional Vocabulary Ontology Library”.And the mutual information method to obtain its emotional category and intensity,thereby constructing an emoji dictionary and a network term dictionary;Considering the words that have modified effects on the expression of emotions and the effects of these modified words on the expression of emotions,this article builds the degree adverbs vocabulary,related words vocabulary,and negative vocabulary basis on the previous studies and the “Synonym Cilin”,and weights are given to degree adverbs vocabulary and related words vocabulary.Then,based on the semantic rules that have been constructed,such as emotional dictionaries,the emotional calculations are performed to obtain the emotional classification and emotional strength of each microblog.Different from the traditional methods based on emotion lexicon and rules,only the word frequency is different.In the classification of sentiment,this paper determines the emotional classification according to the different types of emotional strength,and passes the accuracy,recall value and Value(F)the three indicators compare the two methods.Results:(1)Through the sentiment analysis of the manually labeled microblogs classified as “suspected”,the accuracy rate of the question vocabulary constructed in this paperto identify interrogative emotions is 71.68%,which is significantly higher than the accuracy of common dictionary recognition question emotions 2.51% and that shows the effectiveness of the word dictionary constructed in this paper.(2)Using the accuracy rate,recall rate,and F value to compare the sentiment classification method proposed in this paper and the traditional method of word frequency statistics based on sentiment dictionary,the accuracy and recall rate of the proposed method in each emotional category are better than those based on word frequency.The traditional method of statistics has been improved,which shows the effectiveness of the emotional classification method in this paper.(3)By using the drug-related microblog data as an example,it not only explores the practicality of microblogging fine-grained sentiment analysis,but also provides users with more angles to choose drugs for the same disease so as to support purchase decisions.Conclusion:For the expression of emotional diversities of users in social media,the fine-grainedness of sentiment analysis is an inevitable trend.Because people tend to search,inquire,etc.through the Internet,identifying users' "doubt" emotions is of some significance.In addition,in the process of constructing a sentiment dictionary,this paper fully considers the difference in the expression of different emotional expressions to give the intensity value to the sentiment dictionary through the mutual information method,and combines the semantic rules to calculate the Microblog,and determines the classification based on the intensity.Compared with the traditional method based on word frequency statistics,its accuracy and recall rate have been greatly improved.Not only proved the effectiveness of the proposed method,but also from the actual application,for the user to purchase drugs to provide support.