Bootstrapping Conditional Retrieval for User-to-Item Recommendations

2025-08-22

AI Summary

This paper addresses conditional retrieval in recommender systems—i.e., retrieving items highly relevant to a user under specific conditions (e.g., topics). We propose a dual-tower framework that explicitly incorporates item-side condition information into the query tower representation, enabling joint modeling and deep interaction between user features and conditions—without requiring additional annotated data. The approach supports zero-shot transfer to unseen conditions, ensuring strong generalization and high deployment efficiency. Deployed on Pinterest’s topic-based notification feed, it significantly improves recommendation relevance and drives a 0.26% increase in weekly active users. Our key contribution is the first end-to-end, scalable integration of conditional signals into the dual-tower architecture—balancing effectiveness, inference efficiency, and production practicality.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Computer Vision: Image and Video Retrieval
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

User-to-item retrieval has been an active research area in recommendation system, and two tower models are widely adopted due to model simplicity and serving efficiency. In this work, we focus on a variant called extit{conditional retrieval}, where we expect retrieved items to be relevant to a condition (e.g. topic). We propose a method that uses the same training data as standard two tower models but incorporates item-side information as conditions in query. This allows us to bootstrap new conditional retrieval use cases and encourages feature interactions between user and condition. Experiments show that our method can retrieve highly relevant items and outperforms standard two tower models with filters on engagement metrics. The proposed model is deployed to power a topic-based notification feed at Pinterest and led to +0.26% weekly active users.

Problem

• Enhancing conditional retrieval in recommendation systems
• Incorporating item-side information as query conditions
• Improving relevance and engagement over standard two-tower models

Innovation

• Incorporates item-side information as query conditions
• Bootstraps conditional retrieval using standard training data
• Enables feature interactions between user and condition