Person Versus Non-person Classification of Twitter Handle

Abstract

As marketing based upon the social media profiles is growing at a great pace, finding authenticity of social media accounts is vital for brands. This paper addresses the task of user classification in the micro blogging social media Twitter. We aim to identify whether a Twitter handle is a real person or not. It is done in two steps. First, we segregate human and bot twitter handles to discard the latter. Secondly, we classify whether a human twitter handle is a real person or non-person, e.g., an organization. For the first step, we use a Twitterati identification system [16] which computes various statistical measures from the tweets and use them to segregate human and bot twitter handles. For the second step we use two most widely used and well performing classifiers linear regression (LR) and support vector machine (SVM) for classification of human twitter handles as real person or non-person. We find that SVM outperforms LR. Moreover, the performance of SVM (F1-score = 0.9310) indicates that the proposed method may be used practically in real-life application.