Sony’s ‘FlavorGraph’ Uses AI to Predict How Ingredients Will Combine

AI has entered gaming, autonomous driving, and other areas with mixed success, and now it’s dabbling in cooking. After Google’s AI collided with a Great British Bake Off winner, Sony developed a deep learning system called FlavorGraph designed to combine ingredients like garlic, olives and milk.

Researchers from Sony and Korea University (KU) noted that chefs figured out how to combine ingredients by intuition, resulting in a gradual evolution of pairings like cheese and tomato, pork and apple, and garlic and ginger. Many of these classic combinations were later explained by science, as researchers realized that ingredients sharing dominant flavor molecules often work well together. At the same time, other ingredients that combine well can have very different chemical makeups.


To understand why, the team looked at both molecular information about the ingredients and how they have historically been used in recipes. They then created the FlavorGraph database with flavor profiles such as bitter, fruity, and sweet based on 1,561 flavor molecules. At the same time, they looked at nearly a million recipes to see how the ingredients have been combined in the past.

The resulting data shows the chemicals shared by foods such as wines and citrus groups and how they affect their overall taste, showing which foods might go well with specific wines or fruits. Some of the food pairings in the sample are obvious (cookies and ice cream) and some not (white wine and Campbell’s Condensed Golden Mushroom Soup). Researchers haven’t discovered anything out of the ordinary yet (citing caviar and white chocolate as an example), but the FlavorGraph is just a starting point.

“As science develops and we get better representations of food, we should discover more and more intriguing ingredient combinations, as well as new substitutes for unhealthy or unsustainable ingredients.” , wrote the team.

All products recommended by Engadget are selected by our editorial team, independent of our parent company. Some of our stories include affiliate links. If you buy something through any of these links, we may earn an affiliate commission.

Tech Technology