Want to eat that fantastic Japanese recipe you found on the Internet but with ingredients from the farmers’ market up the street?
Then researchers at the University of Illinois Urbana Champaign have developed an algorithm for you. Professor of Electrical and Computer Engineering Lav Varshney and his team trained a neural network that can substitute recipe ingredients based on what’s available in your part of the world.
“The basic motivating idea is that if one wants to change eating behavior to improve health, then people must want to eat the dishes, and this may be a culturally-defined thing,” Varshney says.
Varshney and his collaborators trained their neural network using nearly 40,000 recipes from the online repository Yummly. They organized the ingredients into 20 pre-selected regional cuisines – like Thai, Russian, or Cajun Creole – and then used a technique called word2vec, common in natural language processing, that locates words in a vector space in order to extract their context.
The program then had the capacity to answer questions like, “What French ingredient corresponds to Japanese soy sauce?”
The researchers hope their new method will lead to healthier eating habits around the world.