Social Media's Toxic Comments Detection Using Artificial Intelligence Techniques

Cyberbullying takes its place in social media and has increased throughout the past few years. The damage that cyberbullying has on the users is undeniable they get attacked either on their appearances, ethnicities, religions, and even their thoughts and personal opinion. The attack causes these users anxiety, depression, low self-esteem, and in the worst scenarios suicide. These harmful actions toward the users drive researchers to identify and detect cyberbullying to fight it. Unfortunately, most of the previous approaches were on English texts, hardly any on other languages. This paper presents a cyberbullying detection system in the Moroccan dialect on an Instagram-collected dataset. The experiment results gave accuracies of around 77% to 91% from both the ML and DL algorithms. The LSTM model gave the best outcome by 91.24% outperforming the ML models.