Research On Microblogging Events With Text Mining

Microblogging has become an important social platform for people to express their opinions and get popular informations.Microblogging often revolves around the same event,which makes these related microblogs have internal relationships.By mining the informations of relevant microblogs of specific event,it is possible to predict the tendency of an event.Nowadays,more and more researchers have studied around microblogging,but only a few researchers pay attention to the microblogging events.As for microblog-ging events,researchers mainly focus on the classification of microblogging events,sentiment classification of microblogging events and related applications.Aiming at the sentimental changes of microblogging events,researchers mainly analyze the rela-tionships between microblogging events and social events.However,the sentimental importance of users and irrelevant sequence mining are neglected.In addition,some microblogs do not contain event tags.If microblogs can be attached with corresponding event tag,we can get more effective informations from corresponding event.For dealing with microblogging events mining and the mentioned problems,the main contributions of this paper are listed as follows:Firstly,we proposed a method of sentiment time series analysis based on mi-croblogging events.Different types of users have different sentiment toward various events.This method aims to study the sentimental importance between different cate-gories of users and events based on the relationships of their sentiment time series.We did some experiments by crawling several events from Sina Weibo.The experimental results showed the sentimental importance of different types of users and the reasons behind them.Furthermore,the reasonableness of this method was also validated based on experimental results.Secondly,in this paper,we proposed a method to mine irrelevant sentimental sub-sequences in sentiment time series.Aiming at the sentiment time series of mi-croblogging events,the irrelevant sentimental sub-sequences are calculated by gener-ating shapelet.We did experiments by crawling multiple events from Sina Weibo.We obtained irrelevant subsequence through experiments and also analyzed the results.In addition,the experiments validated the feasibility of this method.Thirdly,we proposed a method for microblogging event classification based on generative adversarial network.There are some microblogs which are not associated with event tags.Since the text in microblogging are short and their features are sparse,the feature diversity extracted by the existing methods is insufficient.In this paper,we use generative adversarial network to generate diverse features from training data.We did experiments on different event data sets crawled by Sina Weibo and through dif-ferent parameter settings.The experimental results show that this method can achieve better classification performance compared with standard neural networks.