SOCIAL MEDIA HATEFUL CONTENT DETECTION USING DEEP LEARNING RESOURCES

Abstract

Social media platforms enable large-scale communication where users express a wide range of emotions, including anger, frustration, solidarity, satire and political opinions. While these platforms promote interaction and information sharing, it also facilitates the rapid spread of hate speech, offensive language and identity-based hostility. Detecting such harmful content is essential to ensure safe and inclusive digital environments. Existing machine learning approaches rely on hand-crafted textual features, whereas deep learning models automatically learn semantic and contextual representations from data. More recently, transformer-based natural language processing techniques have significantly improved the ability to capture context in social media and code-mixed text. This survey presents a structured and comparative analysis of machine learning; deep learning, transformer-based, hybrid and context-aware approaches for hate speech detection. It examines feature engineering strategies, data characteristics, social media challenges, evaluation methods and interpretability techniques. Furthermore, the survey highlights the importance of multimodal integration, real-time deployment considerations and responsible content moderation systems. Overall, this survey provides a comprehensive review of existing approaches identifies key challenges and outlines emerging research directions in hate speech detection.