Research And Application Of Microblog Events Abstract Generation And Evolution Analysis Technology

With the development of technologies such as social media and mobile Internet,people can use mobile phones to share and interact with information on the social networking site Twitter in real time.This makes Twitter surpass other traditional social media as the first time to publish and disseminate information.When an event occurs,the user will immediately post the relevant tweet on Twitter.Many existing event detection algorithms can easily group these related tweets together,and these tweets are considered as one event.How to detect different stages of development of events dynamically developed with time from a set of related tweets and generate summary presentations to users has become a hot issue to be solved urgently.This thesis proposes a microblogging event summary generation and evolution analysis technique for the Twitter platform.First,it conducts phase detection on Twitter events,then generates a corresponding summary for each important development phase,and finally displays the summary of the different development phases of the event at the time.This thesis focuses on the research and application of summary generation and evolution analysis of microblogging events.The core work is as follows:First,this thesis proposes a detection algorithm based on time window segmentation.According to the fact that the tweets related to events on Twitter are densely distributed,the tweet data is segmented according to the time window,the corresponding representative vector and the representative time are generated for each tweet block,and the tweet blocks are clustered by the clustering algorithm The clustering uses the representative vector and the representative time to calculate the mixed similarity between the tweet blocks.After the clustering is completed,the category containing more tweets is selected as the major development stage of the event.Second,this thesis proposes a hybrid algorithm based on hybrid scoring model.In view of the characteristics of tweeted layers of different tiers on Twitter and different user influence,tweets in each stage of development are scored in turn using a hybrid model,and the tweets with the highest scores are extracted as the summary of the stage of development.The hybrid model considers the tweet summary score based on the undirected graph,the text quality score based on the classifier,and the influence score based on the user's social circle.Experiments show that the algorithm generated by the summary has achieved good results.Thirdly,we design and implement the microblog event summary generation system,and use Twitter data to experiment.The experimental results of two algorithms are given.Experiments show that the detection algorithm based on time window segmentation can improve the accuracy of event detection at the same time,which greatly improves the detection speed of event phase.The summary generated by the microblog event summary generation algorithm based on the mixed scoring model also achieved good results.