Cyberbullying Detection Using Fine-Tuned BERT

Social media platforms have changed the way people communicate, making it easier to share information and build both social and professional connections. Despite these benefits, their extensive use has opened the door to various cybercrimes such as cyberbullying, hate speech, body shaming, trolling etc. These harmful activities negatively impact individuals and society, highlighting the need for advanced and automated detection systems. Therefore, it has become quite important to have efficient and accurate detection methodologies to protect the users of the internet. This paper offers an outlook on identifying cyber bullying through, applying fine-tuned BERT as a classifier of content categorizing it as positive, negative, and neutral. The approach also includes a comparative analysis of some of the cyber-attacks using different BERT family algorithms.