Eco-Amazon: Enriching E-commerce Datasets with Product Carbon Footprint for Sustainable Recommendations
A new dataset uses LLMs to estimate the carbon footprint of 1M+ Amazon products, enabling sustainable AI recommendations.
Researchers Giuseppe Spillo, Allegra De Filippo, Cataldo Musto, and Michela Milano introduced Eco-Amazon, a novel resource that enriches three standard Amazon datasets (Home, Clothing, Electronics) with Product Carbon Footprint (PCF) metadata. They used a zero-shot LLM framework to generate CO2e emission scores for items based on product attributes. This enables developers to build and benchmark the next generation of sustainable recommendation systems that prioritize environmental impact alongside traditional metrics.
Why It Matters
It provides the missing environmental data layer needed to build AI that can actively promote sustainable consumption at scale.