Emergent Event Detectiona And Information Diffusion Modeling On Microblog

Microblogs are beloved by users since its advent because of its timeliness andconvenience, and become users'favorite among the network applications.Microblog becomes a powerful tool for observing the society with its massiveamount of users and relatively free speech. This paper focuses on the detection ofemergent event on microblogs and the modeling of information diffusion throughmicroblog networks, based on the microblog platform.First, we obtain the dataset for microblog study. This paper make aninformation extraction process on "renmin microblog", using the crawler designedaccording to the page information structure of the microblog platform, then extractand store the overall information and hot information of the platform. The overallone-time information extraction includes microblog information, user informationand user relationship information. Hot real-time information extraction includes hotmicroblog information, hot user information and hot key information.Then, we detect the emergent events in microblogs. The briefness of microblogrenders that in its description of the event, the keywords of the events appear in highfrequency, and presented with consistent concern degree of the events. Thus, theemergent event detection in this paper follows the detection of keywords of theevents. First, methods of feature value selection and data organization for microblogcorpus are presented. Then, the feature trajectory of each expected word isconstructed, its time domain and frequency domain feature can judge the burstinessof the word. In this process, the discovery of new words is executed, because thekeywords of the burst events include information of the people and the placeinformation, and these words are usually unknown words but of great significanceto the expression of the events. Finally, the words are clustered according to theconcurrence in microblogs, realizing the correspondence of burst keywords andevents, and the detailed description of the burst event is provided using the obtainedinformation from microblogs.Finally, we construct an information diffusion model for the network ofmicroblog users. This paper constructs a five-tuple model of information diffusionin microblog, combining the route and the characteristics of information diffusion in microblogs, and analyzes the influence factors of the model and the characterizationmethods of elements. Then, simulations are conducted on the process of informationdiffusion in microblogs based on the above analyses. This part focuses on theanalysis of the information communication ability in microblogs, becauseinformation diffusion can be effectively controlled through the control of nodeswith strong information diffusion ability in public opinion supervision.The detection of burst events in this paper reflects the important social eventsauthentically and timely, which helps people to know the current events and enablesthe early warning for relevant institutions. The modeling of information diffusioncan predict the process of information diffusion and corresponding critical nodes,which provides references for monitoring and controlling of information diffusion.