Identificação de discurso de ódio através de algoritmos de Machine Learning

Abstract

The growth of social networks brought as negative factor the spread of hate speech, which impacts society negatively. Due to the large volume of shared comments, manual classification becomes infeasible and the use of computational techniques can assist in the automatic identification of offensive opinions in Portuguese. This article demonstrates the analysis of Supervised Learning methods for rating hate speech in Portuguese. More specifically, techniques already demonstrated in English language research with good results will be employed. The results for F1-Score showed viable the application of techniques such as Mutilayer Perceptrons (80 %) and Random Forest (77 %).