Comparative Analysis of Cyberbullying Detection: A case study for Turkish and English

In this study, we have implemented a significant number of experiments using machine learning (ML), deep learning (DL), and feature extraction techniques to detect cyberbullying in both Turkish and English languages. The results showed that our model performed well using DL with FastText word embedding on both English and Turkish datasets. However, the pre-trained variants of BERT (a language model) performed exceptionally well, achieving an average F-score of 0.9775 for the Turkish dataset and 0.9892 for the English dataset. These findings highlight the effectiveness of using advanced natural language processing techniques for cyberbullying detection in different languages.