TurkHSD: A Hate Speech Detection Model for Turkish Text Content

At <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$a$</tex> time when digital communication is rapidly developing, the spread of hate speech on social media platforms and websites poses a significant problem for social cohesion. In this study, we address the problem of detecting and reducing hate speech in Turkish text content. The study aims to propose a new model to solve the problem by using machine learning methods. We collected and labeled a substantial dataset of 12,989 Turkish text entries from social media and various websites. This data was utilized to train and evaluate multiple machine learning models, including Naive Bayes, Logistic Regression, K-Nearest Neighbors, Decision Tree, Random Forest, and XGBoost. The performances of the models were calculated with known metrics like accuracy, precision, recall, F1 score, and ROC-AUC score. Our findings indicate that Bernoulli Naive Bayes model outperformed other models in terms of accuracy and ROC-AUC score, making it the most effective model for detecting hate speech in Turkish text content. This paper details the methodology, dataset preparation, model training, evaluation, and the implications of deploying such models in real-world applications. Our contributions provide a significant step towards creating safer online environments for Turkish-speaking users by effectively identifying and curbing hate speech.