Any dish worth its salt needs the finest ingredients. In the case of the new Feast cooking app, we needed a structured recipe archive. This is how we rustled it up

You may have heard the Guardian has a new cooking app called Feast. With thousands of searchable recipes, cook mode, and lots of other lovely features, it’s been well received on both iOS and Android. You should definitely get it.

No article on its own could possibly do justice to the work that’s gone into making Feast a reality (a big book might be able to pull it off) but this is a look at a vital piece of the app – its data. There’s no recipe app without recipe data. At least, not one with any functionality worth writing home about.

Structure old recipes with a machine-learning model trained in-house by the thoroughly brilliant data science team.

Drop the resulting data into a temporary recipe curation tool inside which food journalists could use their expertise to refine and approve recipes.

Put the curated data in a database.

In parallel the same functionality could be added to the Guardian’s bespoke content management system, Composer. When it was ready to assume recipe curation responsibilities, we could switch over.

Continue reading...