Dataset or Not? A Study on the Veracity of Semantic Markup for Dataset Pages

Abstract Semantic markup, such as , allows providers on the Web to describe content using a shared controlled vocabulary. This markup is invaluable in enabling broad range of applications, from vertical search engines, rich snippets results, actions emails, many others. In this paper, we focus semantic for datasets, specifically context developing engine datasets Web, Google’s Dataset Search. Search relies identify pages that datasets. While was core technology search, also discovered need address following problem: 61% internet hosts provide do not actually We analyze veracity dataset Search’s Web-scale corpus and categorize where reliable. then propose way drastically increase quality metadata by deep neural-network classifier identifies whether or page with page. Our achieves 96.7% recall at 95% precision point. level enables circumvent noise use high results users.