ParentFull threadlmichelbacher·If I understand the blog correctly, their corpus consists of all the articles they have processed so far. Maybe they have some additional source, e.g. a more general collection of web pages.View on HN