Research On Real-time Computation About The Public Affective Polarity To The Hot Topics On Social Networks

With the development of Internet, hundreds of millions of netizens make theirvoices about hot issues heard through various social platforms such as Weibo andBBS, which producing large amounts of data all the time. With the characteristics ofrapid growth, structural diversity, dynamic updating and wide range, these datacontains various public emotional information of all levels of society. Mining thepublic emotional information makes sense for the research of electronic commerce,information retrieval and public opinion supervision. Mining the public emotionalinformation in social networking has stepped into the stage of processing massive dataso that the demands for real time becomes more. The current way to mine the publicemotional information is similar to a real-time calculation of batch mode, which isnot a true real-time way as well as involves the research for emotional polaritycombining real-time calculation of business and the satisfaction of the features oftexts on the Internet. So this paper studies the real-time mining of public emotionalinformation in social networking from the aspects of text emotional computing andstreaming real-time computing respectively.(1) The accuracy of the text affective computing affects the mining of publicemotional information in social networking directly. The irregular words, complicatedcontext, the Internet terms and rich emotional symbols of public emotionalinformation lead to the more errors in the calculation of the text emotional. Afterstudying the emotion dictionary and the methods for matching the template in thecomputing of text emotions, this paper analyzes the passage of the social networkingby combining the rules matching of the features. Firstly, divide the emotion ontologywords into single polarity words and polarity words, and the former marks theemotional polarity and intensity due to the corresponding emotional dictionary, whilethe latter calculates the emotional polarity by the rules of structure matching and thekeywords matching. Secondly, this paper computes the emotional polarity in thelevels of sentences and passages taking the modifiers, sentences tone and emoticonsinto consideration. (2) Aiming at real-time mining of the public emotions in social networking, andin combination with the characteristics of the social network short text data, this paperproposes a general real-time calculation model about the flow passages, namelyRUBP model. The realization of the core real-time calculation module is based on theTwitter Storm framework. We study the methods of streaming data calculation aboutbusiness mining based on the real-time calculation model, including the ordercalculation and the trend calculation. We also propose the improvements ofdispatching based on the topological structure and traffic to optimize the performanceof RUBP model.(3) This paper carries on several experiments on the real-time calculation of thepublic emotional information in social networking based on the above research. Wecalculate the real-time microblog data of one hot issue by RUBP model and thecalculation method of emotional polarity based on the rules matching. Theexperimental results show the feasibility as well as effectiveness of RUBP model andthe calculation method of the emotional polarity, both of the above research can dowell in real-time calculation in social networks of public emotion polar.