The dataset contains 3 million attribute-value annotations across 1257 unique categories on 2.2 million cleaned Amazon product profiles. It is a large, multi-sourced, diverse dataset for product attribute extraction study. - View it on GitHub