 |
| Sony's FlavorGraph software predicts how the components will be paired |
After competing with Google's artificial intelligence company with the winner of the British competition Bake Off, Sony has developed a deep learning system called FlavorGraph that can be used to link ingredients like garlic, olives, and milk.
Artificial intelligence has entered games, autonomous driving, and other fields with varying degrees of success and is now entering the world of cooking.
In general, FlavorGraph predicts the pairing compatibility of two ingredients by combining information about the molecules of a specific ingredient with how people have used that ingredient in the past.
These suggestions can be used to predict the relationship between compounds and foods. The goal is to develop a smart deep learning model that proposes the use of new supplement and ingredients to help chefs create new foods.
Researchers from Sony and Korea University note that chefs have discovered how to mix different ingredients intuitively, leading to the gradual development of ingredients such as cheese, tomatoes, apples, garlic and garlic. Ginger.
Because researchers realized that ingredients that shared important taste molecules work well together, many of the classic combinations were later explained with science.
At the same time, other well-mixed ingredients can have completely different chemical formulas.
To find out why, the research team examined molecular information about ingredients and how they have been used in recipes in the past.
Therefore, they created a FlavorGraph database of taste characteristics such as bitterness and sweetness based on 1,561 taste molecules.
The team examined nearly a million recipes to see how these ingredients have been added in the past.
The resulting data shows the chemicals shared between foods and how they affect their overall taste, in addition to showing which foods can match a specific type of fruit.
There are clear examples of related foods such as biscuits and ice cream, but some are less clear.
Researchers have found no abnormalities, but FlavorGraph is just the starting point.
The team wrote: As science develops and food metabolism improves, we need to discover increasingly interesting related ingredients, as well as new alternatives to unhealthy or unsustainable ingredients.