Design And Implementation Of Opinion Mining System For Micro-Blog Comments

In recent years,with the rapid development of network technology and the popularity of computers,Internet has become an important medium for people to publish information and express their views.As a representative of the new social media,micro-blog provides convenient space for accumulation and dissemination of public opinion.By micro-blog,users can browse what they are interested in and express their views for others to browse.As a huge social networking platform,there are many hot topics in micro-blog.These hot topics will attract millions of users' attention and comments,which refect the users' attitudes and opinions for the topic.So,opinion analyzing and mining is important for understanding users'view.At present,although opinion mining based on micro-blog's has been widespread concerned and researched,how to do opinon mining based on short-text data in micro-blog still need further research and exploration.Because of openness,anonymity,convenience and other characteristics,micro-blog platform has become the soil for comment spammers growing and spreading.There are a lot of comment spammers which are irrelevant to the hot topic.Those which contain sentiment words will affect the accuracy of opinion mining seriously.Therefore,before opinion mining,this thesis proposes a method to identify comment spammers and remove unrelated data,which can improve the accuracy of view classification for people who comment the topic.Firstly,this thesis analyzes related technologies ablout the micro-blog platform.Protected resource access can be obtained by connecting to micro-blog through OAuth protocol,then micro-blog comments can be obtained based on secondary development and be stored in MySql database.Secondly,this thesis analyzes different comment spammers in micro-blog,and proposes different methods to identify comment spammers,including the dominant comment spammers and hidden comment spammers.Then,an opinion classification model based on SVM is designed and implemented,which can classify the comments data according to sentiment tendency.A prototype system is also designed and implemented,which can analyze the experimental results and show results in tabular form.