Detecting Hate Speech in Portuguese Social Media: A Deep Learning Approach for the Brazilian Presidential Election

Abstract

Social media enables widespread communication with minimal access restrictions, facilitating the expression of emotions and opinions, including harmful content such as hate speech. The dissemination of offensive language online can negatively impact individuals and contribute to social instability. Recent advances in machine learning and natural language processing have improved automated detection of such content. This study focuses on hate speech detection in Portuguese during the 2018 Brazilian presidential elections. We evaluate different deep learning architectures, hyperparameters, and feature configurations to determine optimal parameters for hate speech mining. Additionally, we contribute a Portuguese hate speech lexicon and an annotated dataset, both publicly available on GitHub.