Hotspot Event Dynamic Detection And Analysis On Sina Weibo

With the birth of social network and the vigorous development of various social media,users take social network as a new information-sharing platform to interact and communicate in social media.It is more conviniet to obtain information of interest from social media and to express their relevant attitudes and relevant opinions.At present,the research on detection of hotspot events in online social networks is usually inefficient,and the granularity of the detected events is coarse,which is not easy for users to understand the relevant information of hotspot events.In China,Sina Weibo is an important platform for people to quickly perceive and participate in the discussion of social hotspot events,so how to quickly perceive and detect events from Sina Weibo is of great significance for users to participate in the discussion of events and understand the development tendency of events.For the purpose of dynamic detection of hotspot events and aiming at the deficiencies of existing research,this paper considers using distributed data crawling method to achieve the purpose of real-time collection of microblog data.According to the user's input,this paper collects the original microblog data set related to input keywords from Sina Weibo,and preprocesses the original data set by the hybrid detection method based on semantic correlation and incremental clustering.According to user's different understanding of the event,user can specify the relevant parameters in the event evolution process,such as the time slice size,the number of turning point topics to achieve the purpose of dynamically displaying the event evolution process.The specific work of this paper includes:First,crawling and processing microblog text data.This paper crawls the real microblog data according to the search keywords and expand the data set using expanding keywords.On this basis,this paper designs reasonable strategies to filter the noise data to reduce the influence of meaningless information.Considering the poor normalization of microblog text,this paper preprocesses the microblog text to reduce the impact of useless data and to reduce the difficulty of event detection and analysis process,which can improve the efficiency of event detection and enhance the completeness of event evolution analysis.Second,the event related feature extraction of microblog text.In this paper,based on the preprocessed microblog text data set,the corresponding features are extracted from these preprocessed microblog texts,including entity feature extraction,text semantic feature extraction and keyword weight feature extraction.In the part of entity feature extraction,it mainly includes named entity extraction,part of speech tagging and syntactic relationship feature analysis,which are used to build event tuples;as well as time feature extraction and determination,which helps to build event tuples.These features lay a good foundation for subsequent event detection and analysis task.Third,this paper proposes an incremental clustering event detection method based on hashtag and semantic correlation to detect events from microblog dataset,and realizes event analysis based on chain structure.First,this paper uses the information entropy and event correlation of the hashtag in the original microblog to filter the set of event related hashtags.Then,match the unknown events with the hashtag accurately in one to many ways in using similarity of event related hashtags.Second,according to the event quadruple obtained in the feature extraction stage,conduct incremental clustering based on the hashtag clustering,and finally get the event detection results.According to the event related microblog text set,this paper constructs the chain structure of event tuple,and obtains the evolution process of the event on the time axis.Finally,this paper uses experiments to verify the validity of event detection and analysis method.This paper completes under the background of the cooperation project between the research group and the enterprise.The main work of this paper is the event detection and analysis of Sina Weibo,and the implementation of the corresponding system,which can help network users fully understand the hotspot events on the network,and provide theoretical basis for the public opinion and evolution analysis of hotspot events.