Sentiment Analysis Of Chinese Microblogs Based On The Sentiment Lexicon And The Judgment Of Topic Orientation

With the characteristics such as multiple users,large volume of messages and fast updating,Weibo has become peoples' main way of information acquisition and opinion.Users' opinions on specific topics can be obtained by sentiment analysis of Weibo,which can dig out the implicative value of Weibo text.Through emotional analysis of the content published by Weibo users,users' true emotions can be restored to the greatest degree,which could help the government to control the direction of public opinion,help users optimize their purchasing decisions,and help enterprises to improve market competitiveness by self-improving in a targeted manner.In recent years,emerging online buzzwords have brought great challenges to the sentiment analysis of Weibo.Firstly,a large number of online buzzwords in Weibo text express clear attitudes of Weibo users,while most of the existing sentiment lexicons do not contain these buzzwords.Secondly,the existing word segmentation tools cannot correctly identify online buzzwords,it interferes with word segment and sentence segment of Weibo.Finally,online buzzwords are time-sensitive,so with enhancement or weakening of their popularity,it is necessary to add delete buzzwords from buzzword lexicon in real time.In summary,construct a online buzzword lexicon that can be updated in real time is of great significance to the sentiment analysis of Weibo.What's more,traditional sentiment analysis usually divides the emotions expressed of text into two categories: positive and negative.The existing Chinese sentiment lexicons mainly divides emotional words into positive and negative classes too.Then use various semantic rules to obtain the emotional polarity of Chinese text.However,people's emotional attitudes towards things are often not simple or pure,but complex and diverse.The emotional attitude of users to a Weibo topic should not only stay at the positive and negative levels,but should be subdivided as fine as possible to truly restore human emotions.Therefore,the traditional positive and negative two-category sentiment analysis has not been suitable for the sentiment analysis of Weibo with diverse sentences and diverse vocabulary.Aiming at the above two problems,this paper proposed a method for constructing a network buzzword lexicon that can be updated in real time,and a method for constructing a fine-grained sentiment lexicon based on the semantic similarity calculation method of How Net.Thereby completing the sentiment analysis of Weibo and the fine-grained analysis of Weibo topics.The main work includes the following:(1)Constructed one online buzzword lexicon.Firstly,integrated and analyzed new network word from baidu and Sogou input methods as candidate words for the buzzword lexicon,and then extracted a number of buzzwords by filtering these candidate words in a huge corpus environment.To judge sentimental orientation of these buzzwords,we put forward a Laplacian-based smooth SO-PMI algorithm.Finally the Weibo domain sentiment lexicon including buzzwords was constructed.(2)Emotional analysis of Weibo using the constructed Weibo domain sentiment lexicon.Firstly,according to the lexicon,the emotional polarity of the microblog sentence was calculated,then the Chinese question mark and the period,the modification effect of the linguistic network buzzword are introduced,and finally the position coefficient of Weibo sentence was introduced to obtain the emotional polarity of one whole microblog.(3)Fine-grained sentiment analysis of Weibo topics.By using the semantic similarity calculation method based on How Net,the emotional words were firstly divided into 7 categories.And then defined the emotional intensity from 1 to 9 according to the word similarity.Based on the emotional lexicon of Dalian University of Technology,the fine-grained emotional lexicon of Weibo is constructed.Finally,the Weibo fine-grained lexicon was used to analyze the Chinese Weibo topic orientation.The experimental results show that the online buzzword lexicon constructed in this paper effectively improved the accuracy of clauses and word segmentation,and improved the accuracy of Weibo sentiment analysis.The fine-grained lexicon construction based on the semantic similarity calculation method of How Net can achieve a fine-grained analysis of fine-grained analysis of Weibo topics.