Recipes are more than just lists of ingredients—they’re structured data waiting to be unlocked. In this session, we’ll take you behind the scenes of how we built machine learning models to extract structured data from unstructured recipe articles, enabling the launch of a new recipe application FEAST by the Guardian. We’ll discuss the challenges of working with diverse and inconsistent text data, the techniques we used to train models for ingredient recognition and categorisation, and how we balanced automation with editorial input. Attendees will learn practical insights into using NLP and machine learning for real-world content applications, as well as the impact of these models on user experience and content discovery.
Technical level: High Level/overview
Session Length: 40 minutes
Supported by
