Analysis of FastText with Support Vector Machine for Hate Speech Classification on Twitter Social Media

Hate speech refers to sentences or words that aim to demean or insult individuals, groups, or communities based on factors such as ethnicity, religion, race, or social class. In this study, Natural Language Processing (NLP) techniques were employed using FastText feature extraction and SVM algorithm for text classification. The evaluation was conducted using F1 Score as the performance metric. The data was divided using the Cross-Validation method with 10 folds, and the experiment was performed with four SVM kernels: RBF, Linear, Polynomial, and Sigmoid. The results of this research, based on the effectiveness of the FastTextSVM method combination, demonstrate a strong performance in hate speech classification. By adopting FastText parameters from previous studies and involving four SVM kernels, this research achieved a satisfactory average F1 Score. The results obtained for the Polynomial kernel showed the best performance with an F1 Score of 0.813, followed by the Linear kernel with 0.809, the RBF kernel with 0.808, and the Sigmoid kernel with 0.805. This indicates that the F1 Score results do not show significant differences in outcomes.