Speaker: Gregory Druck, Yummly
Abstract: Although millions of recipes are available online, they typically consist of un- or lightly structured text, which makes organizing, retrieving, and analyzing them difficult. In this talk we discuss methods for structuring and canonicalizing recipes that combine machine learning, manually curated knowledge bases, and user interaction data. We then explore how this recipe representation can be leveraged to surface relevant content and gain insight into how people cook.