Classificação de comunidades em redes temporais por texto: mapeamento dos protestos brasileiros de 2013-2018 no Twitter

Abstract

The decade of 2010 comprised significant social and political changes in Brazil. This thesis aims to revisit some episodes of contentious politics, observed from 2013 to 2018 in the country, in order to analyze virtual communities over time within a social network of wide access and popular use in the country. Social network analysis methods were employed to map controversies within a data set of 5 million tweets during six different episodes: i) the June 2013 protests; ii) the presidential disputes of 2014; iii) the subsequent public demonstrations observed in 2015; iv) 2016; v) 2017; and lastly, vi) the 2018 electoral disputes. The proposed method allowed to evaluate a) structural aspects of the flow of resources shared among members of the social network; b) the similarity among temporal communities, detected by the Leiden method; and c) aspects of groups of greater relevance in each interval. The analysis revealed a decrease in the plurality of communities initially observed, in addition to an increased partisanship of institutional and extra-institutional actors. Furthermore, it is argued that these processes opened space for new political groups, supported by counterpublics and public and private organizations, whose information dissemination on digital platforms was potentially increased by their scale-free structure, as well as the expansion of Internet access in the country over the last decade – specially in urban and densely populated areas – thus reconfiguring the post-2013 Brazilian political scenario.