Social Media Information Dissemination And Emotional Computing

With the rapid development of Internet technology,especially the rise of mobile Internet tools,online social media have become the main way for netizens to disseminate information and express their emotions.Accompanied by the issues of public opinion problem caused by online emergencies such as spread of rumors.How to monitor the spread of information,stop the rumors in time,and guide the social mood correctly are becoming urgent problems at present.This paper mainly deal with the issues of dissemination of information and rumors'classification on social media.In first two chapters we introduce the research status and basic technology related with our methods.Then in the third and the fourth chapters,based on the shortcomings of existing research,we propose below ways for resolving public opinion,stance classification and veracity classification of rumors,the main contribution are as follows:(1)For the research of information dissemination,we propose a multi-agent based simulation method.Based on the complex network,we propose a model of Internet users'willingness to communicate to model the behavior of spreading information of netizens.Then model the agent of government,Internet users and media,design individual behavior and interaction rules.Finally,we use true data prove that our model is useful for modeling the evolution of public opinion.(2)For rumor stance classification task,we mine the features from three dimensions according to the characteristics of tweet corpus,promote the effect of traditional classification method of the task.Then we accord to the traditional method without using the original information and data structure of the rumors,we propose two new models,one is transform the classification problem into sequence labeling issue,process the tree structural data to be sequence samples,use several LSTM-based ways to do the experiments;Another model is proposed by splicing the data in different ways,then use Attention mechanism for classification,our model get a good effect in the relevant data sets.(3)For rumor veracity prediction task,in the condition of limited resources,we use the new feature template and get the best results at present with support vector machine.Then we do some research on the unrestricted resources.