Comparative Analysis of Transformer Based Vector Models for Hate Speech in Turkish Tweets

Hate speech is a rapidly spreading issue on social media platforms and poses a significant problem. This study aims to examine the effects of different sentence transformer models on machine learning algorithms for Turkish hate speech detection. Multilingual-MiniLM, DistilUSE, LaBSE, and DistilBERT, which provide multilingual support, were selected as the sentence transformer models. The study was conducted on a Turkish Twitter dataset created for hate speech detection. The dataset was transformed into vectors using Transformer models and tested with six different machine learning algorithms: Naive Bayes (NB), Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM) and K-Nearest Neighbors (KNN). The evaluation was performed using precision, recall, and F1 score metrics.