Implementação e validação de um sistema protótipo para a detecção de discurso de ódio usando inteligência artificial

Abstract

This work presents the implementation and validation for identifying hate speech in interactive chats, using Python and the pre-trained natural language model BERTimbau. With the growth of digital communication platforms, there has been an increase in disguised language that expresses prejudice and incites violence, often directed at vulnerable individuals or groups. The proposed system aims to detect these manifestations of hate speech and analyze their patterns, offering a more effective approach to content moderation. The practical implementation of the tool demonstrates its applicability in chat environments, contributing to the creation of safer and more respectful interactions. In addition, the study seeks to provide an in-depth understanding of how technology can improve content moderation, protecting users from online abuse. This implementation also investigates the relationship between hate speech and the effectiveness of moderation strategies, contributing to the safety and quality of digital interactions. Through the analysis of collected data, we hope to identify trends and patterns that will help refine moderation techniques and promote a more inclusive digital environment. The use of BERTimbau enhances the accuracy in identifying linguistic nuances, allowing more efficient recognition of harmful speech and opening up new possibilities for the development of protection solutions in chat contexts.