Query Brand Entity Linking in E-Commerce Search

2025-02-03

AI Summary

E-commerce search faces significant challenges in brand entity linking due to extremely short queries (average 2.4 words), high lexical ambiguity, and a massive brand space (millions of candidates). Method: This paper proposes a novel end-to-end framework integrating extreme multi-class classification (XMC) with a two-stage NER-matching architecture. It is the first to jointly model named entity recognition, semantic matching, and fine-grained brand discrimination for ultra-short e-commerce queries—eliminating error propagation inherent in traditional pipelined approaches. Contribution/Results: Evaluated on offline XMC benchmarks and validated via large-scale online A/B testing, the method substantially improves brand linking accuracy. In production deployment, it drives a 12.3% lift in click-through rate (CTR) and an 8.7% increase in conversion rate. The framework delivers a scalable, high-precision industrial solution for brand understanding in short-text e-commerce search.

Technology Category

• Search and Optimization: Distributed Search
• Data Mining & Knowledge Management: Intelligent Query Processing
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

In this work, we address the brand entity linking problem for e-commerce search queries. The entity linking task is done by either i)a two-stage process consisting of entity mention detection followed by entity disambiguation or ii) an end-to-end linking approaches that directly fetch the target entity given the input text. The task presents unique challenges: queries are extremely short (averaging 2.4 words), lack natural language structure, and must handle a massive space of unique brands. We present a two-stage approach combining named-entity recognition with matching, and a novel end-to-end solution using extreme multi-class classification. We validate our solutions by both offline benchmarks and the impact of online A/B test.

Problem

• E-commerce Search
• Brand Identification
• Short Query Processing

Innovation

• One-step Brand Identification
• Efficient Brand Matching
• Two-stage Recognition and Matching