BURSTY TOPIC DETECTION IN TWITTER

Abstract

Twitter has turned out to be one of the biggest microblogging stages for people around the globe to share anything happening around them with companions and past. A bursty subject in Twitter is one that triggers a surge of important tweets inside a brief time of time, which frequently reflects essential occasions of mass intrigue. The most effective method to use Twitter for early location of bursty subjects has accordingly turned into an essential research issue with huge viable esteem. In spite of the abundance of research chip away at point displaying and investigation in Twitter, it remains a test to distinguish bursty themes progressively. Moreover no work has been done in the direction of analysing those bursty tweets. As existing strategies can scarcely scale to deal with the errand with the tweet stream progressively, the Topic Sketch methodology combined (a draw based point show together with an arrangement of methods to accomplish constant recognition)combined with naïve Bayes is proposed to detect as well as to analyse those bursty topics in terms of their polarity or sentiments.The analysis of this method on the tweets comes about with the result that shows both efficiency and effectiveness of this approach. Bursty-Event discovery calculations, have shown that the proposed approach can: (1) accomplish better model execution concerning the assessment criteria; (2) accomplish more exact bursty events on long/short content information.