Research On Event Detection Algorithm For Microblog

Microblog which is known for its convenient way of rapid sharing of information and extensive user connections makes rapid growth and spread of information through the user network exponentially, further exacerbating the data-rich and information-poor contradictions in the information age. From these massive complex microblog data, teasing out the event with the valuable information can not only help users access to information on the event they’re interested in and learn news events going on around them, but also from the perspective of monitoring public opinion and public opinion polls help government departments for emergency management and executive decision.However, as the short, irregular, innovative features for microblog data, using text analysis and data mining technology in the application of traditional network, the effect is no longer ideal. Event detection in particular in the fields of microblog environment faces new challenges, and related research is still in the exploratory phase, urgently seeking for a valid event detecting method for microblog.For the text features of microblog, paper presents the entire event detection model for microblog, and describes the design details of algorithm over every module. Experimental results shows that the paper model can effective, timely and accurately detect the event information for microblog data. The main research contents and innovative points for the paper are as followsFirst, Paper present a new self supervised feature selection method based on n-gram relationship. Different from the traditional model for event detection, paper presents the main part of event detection should be changed from document to features. Representing the microblog features by microblog data itself, the method can express the information better, and also adapt to the characteristics of the microblog.Second, this paper proposes a new word clustering method based on the concept of word activation forces and word affinity. For the new manner, the paper event detection model is successfully converted from document clustering to feature clustering, to associate isolated microblog feature to word clusters in order to express detected events.Third, paper designs and implements a complete and valid microblog event detection model. Based on the innovation of event detection algorithm the thesis puts forward, The paper also gives the corresponding solutions in the area of event detection representation and event detection performance evaluation, providing all aspects of the work of the open ideas for event detection model based on microblog.