Hybrid Moderation Framework for Social Media: Combining AI and Human Expertise
Yayati Nehe, Rushali Sarak, Rutuja Shete, Lata Verma · INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT · 2024
Social media platforms are critical spaces for global communication, allowing users to freely share opinions, engage in discussions, and connect with communities worldwide. However, these platforms are also hosts for harmful content, including hate speech, misinformation, cyberbullying, and offensive imagery. To address these challenges, many platforms implement content moderation systems. This paper provides a comprehensive survey of existing content moderation strategies and proposes a hybrid content moderation system. The hybrid system leverages Natural Language Processing (NLP) for text analysis, Convolutional Neural Networks (CNN) for image detection, and a machine learning model to make final decisions. In cases where the model cannot make a confident decision, human-in- the-loop moderation is invoked. The human decision is final, ensuring fair and transparent content moderation. This paper also explores how each component contributes to improved accuracy and efficiency in detecting harmful content and addresses gaps in handling low-resource languages and multimodal content. Key Words: Social media moderation, hybrid system, NLP, CNN,