CONTEXT-AWARE CONTENT MODERATION USING TRANSFORMER MODELS FOR DETECTING HARMFUL DIGITAL CONTENT

Abstract

The rapid expansion of user-generated content on digital platforms has significantly increased the prevalence of harmful, inappropriate, and offensive material. Traditional content moderation methods, such as rule-based or keyword-based filtering, are increasingly ineffective in identifying subtle forms of harmful content, such as sarcasm, indirect hate speech, or cyberbullying. This paper proposes a novel context-aware model that leverages transformer-based models, such as BERT, RoBERTa, and others, integrated with deep learning techniques to enhance the detection of harmful online content. By focusing on contextual embeddings, our model effectively distinguishes between offensive and non-offensive material, improving the accuracy, precision, and recall of content moderation systems. The experimental results demonstrate that the proposed model outperforms traditional systems and other state-of-the-art models, offering significant improvements in real-time content moderation for large-scale platforms. This research provides a pathway toward more intelligent, fair, and efficient content moderation systems, ensuring safer digital spaces without compromising freedom of expression.