Topic Identification System to Filter Twitter Feeds

Abstract

Twitter is a micro-blogging service where users publish messages of 140 characters. This simple feature makes Twitter the source for concise, instant and interesting information ranging from friends’ updates to breaking news. However, a problem emerge when a user follows many accounts while interested in a subset of its content, which leads to overwhelming tweets he is not interested in receiving. We propose a solution to this problem by filtering incoming tweets based on the user’s interests, which is accomplished through a classifier. The proposed classifier system categorizes tweets into generic classes like Entertainment, Health, Sport, News, Food, Technology and Health. This paper describes the creation and evaluation of the classifier until 89% accuracy obtained.