Automated Conversation Thread Detection and Visualization in WhatsApp Group Chats Using Semantic Clustering

WhatsApp group chats suffer from thread interleaving, yet no annotated datasets exist for this task. This paper presents (1) a synthetic benchmark for WhatsApp thread detection with realistic interleaving patterns, and (2) an automated detection system combining semantic embeddings (Sentence-BERT) with social features (sender patterns, response times) using DBSCAN clustering. Our method achieves NMI=0.463 without requiring prior knowledge of thread count, while baselines requiring this information achieve NMI=0.986 but are impractical for deployment. We demonstrate real-time performance (