TXtract: Taxonomy-Aware Knowledge Extraction for Thousands of Product Categories

Extracting structured knowledge from product profiles is crucial for various\napplications in e-Commerce. State-of-the-art approaches for knowledge\nextraction were each designed for a single category of product, and thus do not\napply to real-life e-Commerce scenarios, which often contain thousands of\ndiverse categories. This paper proposes TXtract, a taxonomy-aware knowledge\nextraction model that applies to thousands of product categories organized in a\nhierarchical taxonomy. Through category conditional self-attention and\nmulti-task learning, our approach is both scalable, as it trains a single model\nfor thousands of categories, and effective, as it extracts category-specific\nattribute values. Experiments on products from a taxonomy with 4,000 categories\nshow that TXtract outperforms state-of-the-art approaches by up to 10% in F1\nand 15% in coverage across all categories.\n