Análise da evolução temporal do discurso de candidatos à prefeitura sobre vacinação contra COVID-19 usando BERTopic

Abstract

The use of social networks, such as Twitter2, has introduced a new dynamic in communication and information flow. This mechanism allows various topics to be discussed simultaneously, and news to spread globally in a short period after publication. This study aimed to apply and evaluate the BERTopic topic modeling approach in the context of short texts about vaccination published on Twitter by mayoral candidates during the 2020 election. The discourse was analyzed based on the topics generated by the model and through the projection of these topics over time. More specifically, the research assessed the evolution of topics by applying diversity, cohesion, and dissonance metrics, comparing the discourse between government and opposition political groups during the pandemic. A comparison was also made with the Latent Dirichlet Allocation (LDA) model developed in a previous study.