Recovery and classification of Twitter user feelings in electoral period

• Autor(a) principal: Matos, Fernanda Fernandes
• Data de Publicação: 2020
• Outros Autores: Magalhães, Lúcia Helena de, Souza, Renato Rocha
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Informação & Informação
• Texto Completo: https://ojs.uel.br/revistas/uel/index.php/informacao/article/view/35310
• topic: Social Networks Analysis; Automatic Classification; Information Retrieval; Information Extraction; Text Mining; Análisis de Redes Sociales; Clasificación Automática; Recuperación de la Información; Extracción de la Información; Minería de Texto; Análise de Redes Sociais; Classificação Automática; Recuperação da Informação; Extração da Informação; Mineração de Texto

Introduction: Social networks have become an important space for users to express their feelings. These comments are valuable for governors to know the citizens 'point of view about their political proposals and candidates perceive voters' reaction to the election campaign. Objectives: To analyze the feelings expressed by users on Twitter, referring to the candidates that competed for the presidency of Brazil in the year 2018, and to predict the result of the elections based on these posts. Methodology: The posts about the candidates that disputed the second round of the elections were the object of study. The software Orange Canvas, a free and open-source machine learning tool, was used to collect the sample and to extract the relevant information. The technique of opinion analysis was applied to automatic classification of feelings into positive, negative and neutral. For better analysis and interpretation of the results, the most important words of comments in word clouds and emotions were shown in frequency distribution charts. Results: There were a lot of negative feelings in the posts and the surprise emotion was the one that stood out the most for both competitors. Conclusions: The study showed that Twitter is an interesting place for users to express their feelings during the election period. However, the work was not able to predict the outcome of the elections based on the emotions. It is believed that this is due to the high rates of rejection of voters regarding the candidates and the polarization that has characterized Brazilian politics in recent times.