Detection of Cyberbullying in Arabic using Machine Learning and ChatGPT

Cyberbullying is one of the complicated problems in Arabic communities. Cyberbullying detection is the main key to control racism and hate speech over the Internet especially on social media. Detection of cyberbullying in Arabic is challenging due to the variety of expressions and different Arabic dialects. Many researches discussed this topic using only one dialect and achieved good results. In this paper, we consolidated datasets with multiple Arabic dialects from different social media platforms. Text pre-processing enhanced the performance of the used models. We classified the comments in the dataset using Naive Bayes model, AraBERT model and ChatGPT for comparison purposes. The experiments showed that AraBERT outperformed the other classifiers for our multi-dialect dataset showing its potential for effectively monitoring Arabic content over the Internet.