Synthetic Data Powers Product Retrieval for Long-tail Knowledge-Intensive Queries in E-commerce Search

2026-02-27

AI Summary

This work addresses the challenge of retrieving relevant items for long-tail, knowledge-intensive queries in e-commerce search, where linguistic diversity, ambiguous intent, and sparse user behavior logs hinder the performance of existing retrieval systems. To overcome these limitations, the authors propose an efficient synthetic data generation framework that leverages a large language model (LLM) with multi-reward signal–guided query rewriting to produce high-quality query–item pairs. These synthetic pairs are integrated into training via an offline high-precision retrieval pipeline, effectively distilling the LLM’s query understanding capabilities into the training data and mitigating distributional shift. Remarkably, incorporating only this synthetic data yields substantial gains in recall relevance, with online side-by-side human evaluations confirming a noticeable improvement in user search experience.

Technology Category

• Search and Optimization: Learning to Search
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Product retrieval is the backbone of e-commerce search: for each user query, it identifies a high-recall candidate set from billions of items, laying the foundation for high-quality ranking and user experience. Despite extensive optimization for mainstream queries, existing systems still struggle with long-tail queries, especially knowledge-intensive ones. These queries exhibit diverse linguistic patterns, often lack explicit purchase intent, and require domain-specific knowledge reasoning for accurate interpretation. They also suffer from a shortage of reliable behavioral logs, which makes such queries a persistent challenge for retrieval optimization. To address these issues, we propose an efficient data synthesis framework tailored to retrieval involving long-tail, knowledge-intensive queries. The key idea is to implicitly distill the capabilities of a powerful offline query-rewriting model into an efficient online retrieval system. Leveraging the strong language understanding of LLMs, we train a multi-candidate query rewriting model with multiple reward signals and capture its rewriting capability in well-curated query-product pairs through a powerful offline retrieval pipeline. This design mitigates distributional shift in rewritten queries, which might otherwise limit incremental recall or introduce irrelevant products. Experiments demonstrate that without any additional tricks, simply incorporating this synthetic data into retrieval model training leads to significant improvements. Online Side-By-Side (SBS) human evaluation results indicate a notable enhancement in user search experience.

Problem

• long-tail queries
• knowledge-intensive queries
• product retrieval
• e-commerce search
• behavioral logs

Innovation

• synthetic data
• query rewriting
• long-tail queries
• knowledge-intensive retrieval
• LLM distillation