Brazilian Presidential Elections in the Era of Misinformation: A Machine Learning Approach to Analyse Fake News

Abstract

As Brazil faced one of its most important elections in recent times, the fact-checking agencies handled the same kind of misinformation that has attacked voting in the US. However, stopping fake content before it goes viral remains an intense challenge. This paper examines a sample database of the 2018 Brazilian election articles shared by Brazilians over social media platforms. We evaluated three different configuration of Long Short-Term Memory. Experiment results indicate that the 3-layer Deep BiLSTMs with trainable word embeddings configuration was the best structure for fake news detection. We noticed that the developments in deep learning could potentially benefit fake news research.