Long-term Characterization of Political Communications on Social Media

Social media play an important role in shaping political discourse, creating a public sphere that enables discussions, debates, and deliberations. Therefore, several studies analyzed and characterized the communication of politicians in social media through political and non-political short text messages. However, classifying short messages as political is not a trivial task, especially considering that political topics change regularly over time. While some politicians only share professional communications about their political agenda and activities, others prefer a more personal and informal approach, sharing communications about the most varied subjects, such as religion, sports, and their families. Thus, in this work, we propose a methodology to characterize the communication of Brazilian politicians over the years in terms of the amount of political and non-political messages they post. For this task, we collected more than 3M tweets from 914 Brazilian deputies from October 2013 to October 2019 (6 years span). We characterize their communications through a machine learning classifier that labels all tweets as political and non-political and is robust to concept drifts. We found that the main concept drifts occurred during important events of Brazilian politics, such as the impeachment of former president Dilma Rousseff in 2016 and presidential and municipal elections. Moreover, we characterize the explosive rise of the right seen just before the 2018 elections. Although left-wing politicians post more on social media, the participation of right-wing politicians is broader and more evenly distributed. Finally, we reveal the increase of public engagement over time, especially with the left and right political tweets.