Boosting Product Discovery in e-Commerce Search with Related Intent Generation

Ji Xin, Xiao Xiao, Ishan Bhatt, Vinesh Gudla, Trace Levinson, Raochuan Fan, Shishir Kumar Prasad, Prakash Putta, Tejaswi Tenneti · July 29, 2026

Traditional e-commerce search engines focus on finding exact matches to a user's query, which can miss out on relevant substitute or complementary products. This paper from Amazon scientists introduces a system that expands search results by generating related user intents—like 'dinner party appetizers' for a 'chips' query. The system uses a two-stage approach: first, a large language model (LLM) generates intents for popular queries; then, a smaller, fine-tuned model (SLM) extends this to rare queries via knowledge distillation. This boosts discovery coverage from 60% to 80% of queries at only 30% of the LLM's cost. The method not only improves user satisfaction but also helps surface long-tail and emerging products, potentially acting as a marketplace balancing mechanism. For investors, this means better user engagement and increased sales opportunities for a wider range of products.

Read more: https://arxiv.org/abs/2607.27172

This is an AI-generated summary, not a reproduction of the original. Read it at arXiv.

Read the full research at arXiv →

Try TreasureX free →

Educational content only, generated with AI assistance from public research abstracts. Not investment advice. Thank you to arXiv for use of its open access interoperability.

🔒

This is a Pro feature

Read the full research digest article with a Pro subscription.

Upgrade to Pro