Exploiting Emojis in Detection of Cyberbullying in Arabic Social Media

Abstract

With the lack of laws and regulations on people's behavior on social media platforms, people tend to express their honest feelings toward others and trending news. Emojis are used extensively to express feelings and in many cases, comments can also include emojis. Cyberbullying is a major issue in almost all social media platforms all over the world, which includes the Arab world. Previous studies have largely focused on text-only approaches, with limited research incorporating emoji analysis for cyberbullying detection in Arabic social media. This research aims to investigate the effect of emojis in cyberbullying detection in Arabic comments on Instagram. Long Short-Term Memory (LSTM) neural networks and Support Vector Machines (SVMs) were used to detect cyberbullying in Arabic content. The performance of the models was studied with and without the inclusion of emojis in the comments. Simulation results shows that the consideration of emojis in developing detection models improved the accuracy by 2% and F1 score values by 0.15 compared to text-only models. This work considers the importance of cultural-specific expressions, including emojis, in developing more effective content moderation systems for social media platforms.