Ai Powered Content Moderation And Alert System

Abstract

Social Media Platforms Such As Twitter (now X) Have Become Primary Channels For Real-time Information Exchange, En- Abling Rapid Dissemination Of News, Opinions, And Public Discourse. However, This Rapid Growth Has Also Led To A Significant Rise In Spam, Misinformation, Abusive Language, And Malicious Content, Which Negatively Impact User Experience And Platform Credibility. Manual Moderation Methods Are No Longer Effective Due To The Massive Volume And Velocity Of Incoming Tweets, Making Automated Content Moderation Systems Essential. This Survey Examines Existing Machine Learning, Deep Learning, And Transformer-based Techniques For De- Tecting Spam And Harmful Content On Twitter. It Begins With Tradi- Tional Text Classification Approaches And Progresses Toward Advanced Transformer Models That Provide Improved Semantic Understanding, Such As Sentence-BERT (SBERT). Particular Emphasis Is Placed On Hybrid Deep Learning Architectures That Combine SBERT-based Semantic Embeddings With Convolutional Neural Networks (CNN) And Long Short-Term Memory (LSTM) Networks. These Hybrid Models Aim To Capture Both Local Textual Patterns And Long-range Contextual Dependencies Present In Tweet Streams. The Survey Also Highlights The Importance Of Severity-based Content Classification, Where Detected Content Is Categorized According To Risk Levels To Support Real- Time Alerts And Administrative Monitoring. A Comparative Analysis Of Existing Approaches Reveals Performance Limitations Related To Scalability, Generalization, And Real-time Applicability. While Hybrid Deep Learning Models Demonstrate Promising Results In Multi-class Classification Tasks, Several Research Challenges Remain Unresolved. The Study Concludes By Identifying Future Directions Toward Developing Scalable, Accurate, And Practical Twitter Content Moderation Systems Capable Of Adapting To Evolving Online Behaviors.

Keywords