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

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8489–8502 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 8489 TXtract: Taxonomy-Aware Knowledge Extraction for Thousands of Product Categories Giannis Karamanolakis∗ Columbia University New York, NY 10027, USA gkaraman@cs.columbia.edu Jun Ma, Xin Luna Dong Amazon.com Seattle, WA 98109, USA {junmaa, lunadong}@amazon.com Abstract Extracting structured knowledge from product profiles is crucial for various applications in e-Commerce. State-of-the-art approaches for knowledge extraction were each designed for a single category of product, and thus do not ap- ply to real-life e-Commerce scenarios, which often contain thousands of diverse categories. This paper proposes TXtract, a taxonomy- aware knowledge extraction model that applies to thousands of product categories organized in a hierarchical taxonomy. Through cate- gory conditional self-attention and multi-task learning, our approach is both scalable, as it trains a single model for thousands of cate- gories, and effective, as it extracts category- specific attribute values. Experiments on prod- ucts from a taxonomy with 4,000 categories show that TXtract outperforms state-of-the-art approaches by up to 10% in F1 and 15% in coverage across all categories. 1 Introduction Real-world e-Commerce platforms contain bil- lions of products from thousands of different cat- egories, organized in hierarchical taxonomies (see Figure 1). Knowledge about products can be rep- resented in structured form as a catalog of prod- uct attributes (e.g., flavor) and their values (e.g., “strawberry”). Understanding precise values of product attributes is crucial for many applica- tions including product search, recommendation, and question answering. However, structured at- tributes in product catalogs are often sparse, lead- ing to unsatisfactory search results and various kinds of defects. Thus, it is invaluable if such structured information can be extracted from prod- uct profiles such as product titles and descriptions. Consider for instance the “Ice Cream” product of Figure 1. The corresponding title can potentially ∗Work performed during internship at Amazon. … Product Groceries Health … Fruit Alcoholic Beverages Beer Sports Nutrition … Protein Powder Beauty … Hair Care Shampoos Lager Ice Cream Ben & Jerry's Strawberry Cheesecake Ice Cream 16 oz … • Great idea for a delicious dessert • Includes Fairtrade certified sugar • Kosher certified dairy Figure 1: A hierarchical taxonomy with various prod- uct categories and the public webpage of a product as- signed to “Ice Cream” category. be used to extract values for attributes, such as “Ben & Jerry’s” for brand, “Strawberry Cheese- cake” for flavor, and “16 oz” for capacity. State-of-the-art approaches for attribute value extraction (Zheng et al., 2018; Xu et al., 2019; Rezk et al., 2019) have employed deep learning to capture features of product attributes effectively for the extraction purpose. However, they are all designed without considering the product cat- egories and thus cannot effectively capture the di- versity of categories across the product taxonomy. Categories can be substantially different in terms of applicable attributes (e.g., a “Camera” product should not have flavor), attribute values (e.g., “Vi- tamin” products may have “fruit” flavor but “Ba- nana” products should …

Content truncated to 500 words