Real-Time Voice-Based Cyberbullying Detection Using AI in Social Media and Gaming

Abstract

The research proposes developing a real- time AI system that detects cyberbullying when it happens over voice chats in online games and on various social media networks. The proposed method here uses the Speech Recognition which removes the noise and disturbance in the audio to make speech conversion into text form more accurate. Then TF-IDF system is used to detect major terms and give extra attention to the most important information from the text. The model was examined to make sure it runs smoothly and fast when it is actually being used in real time. This approach is more accurate, faster and reliable in a different factor. After performing many tests, the LinearSVC has exhibited high detection accuracy, F1 scores, and less false positives. The main advantage of LinearSVC is that it can handle both small and large learning datasets effectively. As a result, the algorithm is more accurate, faster and performs stably than others, so it is helpful for real applications because it provides strong, fast and scalable safety.