Classification of Text, Image and Audio Messages Used for Cyberbulling on Social Medias

Cyberbullying has become an increasingly significant cultural issue in recent years. A person is affected by cyberbullying in both psychological and emotional ways. Bullying that takes place via electronic devices, such as a computer, a smartphone, or a tablet computer, is known as cyberbullying. Harassment in the digital world can take many forms, including but not limited to transmitting or uploading insulting, harmful, inaccurate, or offensive material about another person via text, message, or program; or on social networking sites, message boards, and online games. Disclosure of personal or sensitive information about another individual might be considered cyberstalking. This paper will analyze several media types (text, images, and videos) posted on social media with the goal of identifying instances of cyberbullying. In this study, a graph convolutional neural network, a pretrained Googlenet, a Mel-scale filter bank speech spectrogram, and a CNN network model are introduced for use in audio post-classification. The study’s primary findings indicate that audio post-processing with MFCC’s and graph convolutional neural networks generates improved outcomes, including one-dimensional representation, for both text and image properties. To this end, we used a combination of GCN and Melfrequency cepstrum to represent text, image, and video input in this setting, with a resulting 85% accuracy of a bullying class.