A Novel Method for Clustering Words in Micro-Blogs Texts and its Application to Event Discovery

This paper exhibits a novel method for clustering words in small scale online journals, in light of the likeness of the related fleeting arrangement. Our method, named SAX*, utilizes the Symbolic Aggregate ApproXimation calculation to discredited the fleeting arrangement of terms into a little arrangement of levels, prompting a string for each and then characterize a subset of "fascinating" strings, i.e. those speaking to examples of aggregate consideration. Sliding worldly windows are utilized to distinguish co-happening groups of tokens with the same or comparative string. To survey the execution of the method, first tune the model parameters on a 2-month 1 % Twitter stream, amid which various around the world occasions of contrasting sort and length (sports, legislative issues, calamities, wellbeing, and famous people) happened. At that point, assess the nature of every single found occasion in a 1-year stream, "goggling" with the most successive bunch n-grams and physically surveying what number of bunches compare to distributed news in a similar fleeting space. At long last, play out a unpredictability assessment and contrast SAX* and three alternative methods for occasion revelation. Our assessment demonstrates that SAX* is no less than one request of extent less complex than other fleeting and non-transient ways to deal with smaller scale blog bunching.