ADORE: Autonomous Domain-Oriented Relevance Engine for E-commerce

TL;DR

ADORE is a novel framework that enhances e-commerce relevance modeling by integrating rule-aware discrimination, adversarial data synthesis, and domain-specific knowledge distillation, improving robustness and efficiency.

Contribution

It introduces a self-sustaining, resource-efficient relevance engine combining multiple innovations to address semantic gaps and data scarcity in e-commerce search relevance modeling.

Findings

Effective in large-scale experiments

Improved relevance accuracy in online A/B tests

Enhanced robustness against adversarial examples

Abstract

Relevance modeling in e-commerce search remains challenged by semantic gaps in term-matching methods (e.g., BM25) and neural models' reliance on the scarcity of domain-specific hard samples. We propose ADORE, a self-sustaining framework that synergizes three innovations: (1) A Rule-aware Relevance Discrimination module, where a Chain-of-Thought LLM generates intent-aligned training data, refined via Kahneman-Tversky Optimization (KTO) to align with user behavior; (2) An Error-type-aware Data Synthesis module that auto-generates adversarial examples to harden robustness; and (3) A Key-attribute-enhanced Knowledge Distillation module that injects domain-specific attribute hierarchies into a deployable student model. ADORE automates annotation, adversarial generation, and distillation, overcoming data scarcity while enhancing reasoning. Large-scale experiments and online A/B testing verify the effectiveness of ADORE. The framework establishes a new paradigm for resource-efficient, cognitively aligned relevance modeling in industrial applications.