Enhancing Ad Relevance: Integrating Real-Time Context into Sequential Recommender Models
Pinterest engineers integrated real-time context into sequential recommender models to enhance ad relevance, particularly on the Related Pins surface. This was achieved through a new Contextual Sequential Two Tower Model architecture, which incorporates a context layer into the query tower and uses synthetic augmented data to learn from real-time context during offline training. The model demonstrated a 3x to 10x increase in Recall@K and a 275-300% increase in candidate median relevance, resulting in a 0.7% lift in conversion-related ROAS.
Machine LearningData