Linguistic profiles on microblogging platforms to characterize political leaders: The Ecuadorian case on Twitter

Social interaction on microblogging platforms is becoming a reliable instrument for studying political communication characteristics. Microblogging platforms, such as Twitter, let citizens to engage in the political debate generating well-defined profiles in the platform. Using publicly available tweets it is possible to build a linguistic profile to compare leaders and average citizens. We describe the linguistic analysis of 330,000 tweets collected from 221 Ecuadorian tweeters classified into three different profiles: political leaders, leaders' followers, and average local users. We build a feature vector for each user's tweets using 12 psychological dimensions included in the LIWC (Linguistic Inquiry Word Count) text analysis software and compare users with different profiles using those vectors. Our findings show that the leaders group exhibits a different linguistic profile from the others two groups: around 30% of leader followers are similar to at least one leader while just 19% of average local users are similar to at least one leader. Furthermore, the results of our analysis allows to determine whether local users have some similar characteristics of language uses on social networks of political leaders' followers without relying on critical discourse analysis.