Real-Time Cyberbullying Detection Across Multimodal Platforms Using XLNET and Deep CNN

Cyberbullying has emerged as a primary societal concern with the widespread use of social media and online communication systems. The developing complexity of virtual interactions, together with text messages, social media comments, and images with embedded textual content, makes computerized detection way more challenging and complex. Conventional methods frequently fail to adequately convey the complex and dynamic character of cyberbullying practices. Here, we recommend a robust solution which makes use of Deep Convolutional Neural Networks (Deep CNNs) for potent classification of messages and XLNet for feature extraction. This can process a variety of input types, including textual and visual information, to enhance the identification of bullying patterns more efficient by integrating these contemporary system learning techniques. This framework's ability to detect various forms of online movement makes it more effective across multiple modalities thus contributing to the development of a safer and more inclusive online environment through corrective action for all users.