Longitudinal analysis of brazilian politicians’ discourse: a topic approach

Abstract

The exposure of corruption scandals, as well as the political polarization in Brazil since the 2018 elections, installed a sharp and deep political and social crisis in the country. These conflicts were also present on social media, which set the stage for heated discussions between politicians and the general public. In this context, our work aims to use computational methods to analyze social media data, in order to identify topics of political messages posted by Brazilian deputies from 2013 to 2019. For this task, we adapted the biterm topic model (BTM ), an algorithm that finds topic models in short texts, to enable topic analysis over time. From this model, we investigated the behavior of politicians on social networks (Twitter), identifying the main issues discussed over time. Our approach divides the studied period into annual segments and compares the topic models from different time intervals, building a topic similarity graph. Regarding the investigation of the evolution of temporal topics, our results showed that themes related to the political crisis and activities related to the Chamber of Deputies were the most discussed by deputies during the period studied.