Publicação

Predicting an election’s outcome using sentiment analysis

Ver documento

Detalhes bibliográficos
Resumo:Political debate - in its essence - carries a robust emotional charge, and social media have become a vast arena for voters to disseminate and discuss the ideas proposed by candidates. The Brazilian presidential elections of 2018 were marked by a high level of polarization, making the discussion of the candidates’ ideas an ideological battlefield, full of accusations and verbal aggression, creating an excellent source for sentiment analysis. In this paper, we analyze the emotions of the tweets posted about the presidential candidates of Brazil on Twitter, so that it was possible to identify the emotional profile of the adherents of each of the leading candidates, and thus to discern which emotions had the strongest effects upon the election results. Also, we created a model using sentiment analysis and machine learning, which predicted with a correlation of 0.90 the final result of the election.
Autores principais:Martins, Ricardo
Outros Autores:Almeida, J. J.; Henriques, Pedro Rangel; Novais, Paulo
Assunto:Emotion analysis Machine learning Natural processing language Sentiment analysis
Ano:2020
País:Portugal
Tipo de documento:comunicação em conferência
Tipo de acesso:acesso aberto
Instituição associada:Universidade do Minho
Idioma:inglês
Origem:RepositóriUM - Universidade do Minho
Descrição
Resumo:Political debate - in its essence - carries a robust emotional charge, and social media have become a vast arena for voters to disseminate and discuss the ideas proposed by candidates. The Brazilian presidential elections of 2018 were marked by a high level of polarization, making the discussion of the candidates’ ideas an ideological battlefield, full of accusations and verbal aggression, creating an excellent source for sentiment analysis. In this paper, we analyze the emotions of the tweets posted about the presidential candidates of Brazil on Twitter, so that it was possible to identify the emotional profile of the adherents of each of the leading candidates, and thus to discern which emotions had the strongest effects upon the election results. Also, we created a model using sentiment analysis and machine learning, which predicted with a correlation of 0.90 the final result of the election.