Detecting Kids Cyberbullying Using Transfer Learning Approach: Transformer Fine-Tuning Models

The rising popularity and the usage of social media have given easy access to social media apps such as Facebook, Twitter, Instagram, and YouTube. Numerous entertaining and educative movies and videos are created and uploaded daily to attract a young audience, particularly kids and teenagers. This has caused a significant rise in inappropriate content upload and cyberbullying. The comments received on such videos can adversely impact the mental state, and behaviour child’s if not monitored by the parents. There are researchers done to solve the cyberbullying issue through the classical machine, deep learning, and natural language processing approaches. However, such methods require computational time and hardware while consuming a massive amount of data to train models to improve detection accuracy. This chapter proposes a method to detect cyberbullying on online platforms using the method of transfer learning. Several experiments were carried out using the existing dataset and a variant of pre-trained models based on BERT. Experiment results show that transfer learning outperformed classical machine, deep learning, and learning methods.