Search papers, labs, and topics across Lattice.
This paper presents a production-scale system for topical trend detection on Snapchat, leveraging multimodal topic extraction from short-form videos, time-series burst detection, and LLM-based consolidation and enrichment. The system identifies emerging trends by clustering video content, detecting bursts in topic frequency, and using LLMs to refine and contextualize these trends. Offline human evaluation over six months confirms high precision, and deployment has demonstrably improved content freshness and user experience.
Snapchat's new trend detection system proves that LLMs can successfully consolidate multimodal signals at scale to surface emerging topics from short-form video, boosting content freshness and user engagement.
Automatic detection of topical trends at scale is both challenging and essential for maintaining a dynamic content ecosystem on social media platforms. In this work, we present a large-scale system for identifying emerging topical trends on Snapchat, one of the world's largest short-video social platforms. Our system integrates multimodal topic extraction, time-series burst detection, and LLM-based consolidation and enrichment to enable accurate and timely trend discovery. To the best of our knowledge, this is the first published end-to-end system for topical trend detection on short-video platforms at production scale. Continuous offline human evaluation over six months demonstrates high precision in identifying meaningful trends. The system has been deployed in production at global scale and applied to downstream surfaces including content ranking and search, driving measurable improvements in content freshness and user experience.
