AI-PAVE-Br: Leveraging Large Language Models for Enhanced Product Attribute Value Extraction through a Golden Set Approach

2026-06-23

AI Summary

This study addresses the low accuracy of attribute extraction from Portuguese-language e-commerce product descriptions in Brazil, a challenge stemming from linguistic complexity and category diversity. To tackle this, the authors propose AI-PAVE-Br, a system that leverages large language models enhanced with domain-specific prompt engineering and a three-tier structured annotation schema—spanning entities, categories, and subcategories—tailored to e-commerce contexts. The work contributes the first high-quality, manually annotated Portuguese e-commerce dataset, “Golden Set,” on which AI-PAVE-Br demonstrates significantly superior performance over conventional named entity recognition approaches. By delivering a scalable and accurate solution for structured information extraction in non-English markets and publicly releasing the dataset, this research advances both practical applications and future scholarly inquiry in multilingual e-commerce NLP.

Abstract

The explosive growth and complexity of product data within the dynamic Brazilian e-commerce landscape demand robust and specialized methods for structured information extraction. Traditional approaches to Product Attribute Value Extraction (PAVE) often struggle with the linguistic nuances and sheer diversity of product descriptions in Portuguese. To address this critical gap, this paper introduces two major contributions. First, we present AI-PAVEBr, a specialized system engineered with Large Language Models (LLMs) to perform high-accuracy PAVE specifically for Brazilian e-commerce catalogs. Second, to facilitate reproducible research and provide a definitive benchmark, we introduce and share the Golden Set, a new, meticulously curated, and manually annotated dataset for PAVE in Portuguese. We detail the creation process and structure (Entity, Category, Subcategories) of this high-quality reference set. Our experiments conclusively show that AI-PAVE-Br, leveraging targeted prompt engineering, dramatically outperforms conventional Named Entity Recognition (NER) baselines. This work not only delivers a superior, scalable solution for a major non-English market but also enriches the NLP community with a valuable, publicly available resource for future PAVE research.

Problem

• Product Attribute Value Extraction
• Brazilian e-commerce
• Portuguese language
• structured information extraction
• linguistic nuances

Innovation

• Large Language Models
• Product Attribute Value Extraction
• Prompt Engineering
• Golden Set
• Portuguese NLP