IDENTIFICATION OF HATE SPEECH IN THE 2020 MUNICIPAL ELECTIONS

Hate speech has become increasingly prevalent on social media, with evidence suggesting that it intensifies during election years. It is frequently a topic of discussion in public policy debates, with the aim of holding perpetrators accountable and mitigating its effects without infringing upon individual freedom of expression. In this context, artificial intelligence can be a valuable tool to aid in the identification of hate speech. This study presents the construction and evaluation of a Naïve Bayes classifier to identify hate speech in publications on social media platform X related to candidates who contested the second round of the 2020 municipal elections. A quantitative methodology was employed. The CRISP-DM process, well-established in the field of data science, was adopted for the construction and evaluation of the proposed Naïve Bayes classifier. The data used to train the classifier was collected through an exploratory search on Kaggle, an online repository of data for training predictive models. The data used to evaluate the model was collected from a survey of publications on social media platform X. The classifier's performance was evaluated quantitatively using statistical accuracy metrics. Overall, the proposed Naïve Bayes classifier achieved an average accuracy of 72.38% on the dataset collected for evaluation. The performance difference between the proposed classifier and Perspective API, an online tool for hate speech identification adopted as a reference in this study, was less than 9.12% for all candidates considered. These results demonstrate that the proposed Naïve Bayes classifier was capable of identifying the presence of hate speech in publications on social media platform X related to candidates who contested the second round of the 2020 municipal elections.