BULLY NET: UNMASKING CYBER BULLIES ON SOCIAL NETWORKS

Abstract

In the rapidly evolving landscape of online communication, the surge in cyberbullying has emerged as a critical challenge, necessitating innovative solutions for detection and prevention. Existing approaches often reply on simplistic keyword-based filters or rule-based methods, struggling to keep pace with the dynamic nature of cyberbullying scenarios. The intricate and varied nature of online harassment demands a more sophisticated system capable of discerning subtle nuances within social media interactions. Recognizing this gap, the proposed BullyNet system introduces a character-level convolutional neural network (CNN) approach to enhance the accuracy and adaptability of cyberbullying detection. By incorporating both word-based and character-based models, BullyNet aims to provide a holistic understanding of language expression and contextual cues, offering a nuanced solution to the complex challenges posed by cyberbullying. This system's multifaceted approach, encompassing preprocessing, training, and evaluation of CNN models, is designed to address the shortcomings of existing systems and contribute to the creation of a safer online environment. BullyNet stands as a promising stride towards unmasking cyberbullies on social networks, emphasizing the need for advanced tools capable of navigating the intricate landscape of digital communication.