Detecting Cyberbullying on Social Networks Using Language Learning Model

This paper introduces an advanced system designed to analyze and classify potential cyberbullying content in images from Twitter. Utilizing the image processing of OpenAI's GPT-4, the system generates description metadata for each image. This data is then stored and managed within a MongoDB database, setting the stage for the subsequent analytical phase. We employed Language Learning Model (LLM) to examine the AI-generated image descriptions, assessing them for indicators of cyberbullying. Through AI technologies, the system highlights the potential of integrating AI to address critical social issues in digital communication platforms. Finally, we presented the preliminary experimental results to substantiate the potential of our proposed approach.