Content-Based Discovery of Twitter Influencers Chiara Francalanci, Irma Metra Department of Electronics, Information and Bioengineering Polytechnic of Milan, Italy irma.metra@mail.polimi.it chiara.francalanci@polimi.it Abstract Identifying social media influencers in a given domain is considered key to building a brand’s reputation. Influencers are opinion makers who play a critical role in determining the dynamics with which information spreads across a social network. In Twitter, a large number of followers is considered a fundamental indicator to discover influencers. The assumption is that a user with a large number of followers has a large audience and, thus, is more likely to influence the opinion of people in any given domain. Our claim is that influencers can exert an influence only when the content that they share is considered interesting by their followers. In this paper, we propose a content-based measure of influence, called COAX that includes, but is not limited to the number of followers. COAX is tested on a sample of over 10.000 users from random domains according to the Analytic Hierarchy Process (AHP). Preliminary results show how COAX can provide a ranking that is significantly different from that obtained by means of the number of followers alone. Keywords: social media; influencers; influence; Twitter. 1. Introduction Identifying social media influencers in a given domain is considered key to building a brand’s reputation (Bruni L. , 2014). Influencers are opinion makers who play a critical role in determining the dynamics with which information spreads across a social network. In Twitter, a large number of followers is considered a fundamental indicator to discover influencers. The assumption is that a user with a large number of followers has a large audience and, thus, is more likely to influence the opinion of people in any given domain. Our claim is that influencers can exert an influence only when the content that they share is considered interesting by their followers. As a consequence, they are influential in selected domains where they have the capability to share interesting content. The previous academic literature supports our claim by showing how a variety of variables describing content can have an impact on the probability with which content itself is shared. For example, in (Bruni, Francalanci, & Giacomazzi, 2013) the authors claim that linking multimedia content in a Tweet increases the average number of retweets. In (Boyd, Golde, & Lotan, 2010) authors note that a content that has had an impact on a user’s mind is shared. In (Suh, Hong, Pirolli, & Chi, 2010) authors discover how most content is retweeted only once and (Ota, Maruyama, & Terada, 2012) introduces the concept of depth of retweets to measure the impact of the original tweet. Overall, the academic literature is heading towards the concept of influence, i.e. the actual impact that a tweeter has on his audience and on other users that they are not directly connected with.