Artificial intelligence that can predict taste from beer's chemical composition could help create alcohol-free versions that taste exactly like regular beer.
Predicting flavor from a compound is difficult because complex interactions between ingredients and the psychology of taste can result in surprisingly different perceptions, even among people who have tasted the same thing.
To deal with this, kevin verstrapen Professors at the University of Leuven in Belgium have developed an AI model that can predict the flavor profile of beer based on its chemical composition and suggest ways to improve the flavor.
The model was trained on beer reviews by a panel of 16 expert tasters who scored each beer on 50 attributes, as well as 180,000 public ratings from online beer rating websites. The study compared these subjective descriptions with measurements of 226 compounds in 250 Belgian beers.
“Hundreds of these compounds are received by our nose and mouth, but mostly in the nose, and then processed in the brain to give us what we think of as flavor,” Verstrepen says. “The fact that we can actually predict this accurately using machine learning is pretty amazing.”
Verstrepen and his team used this model to predict how to improve the flavor of beer by adding mixtures of specific compounds, such as lactic acid and ethyl acetate. The resulting beer was given significantly higher ratings by a panel of trained tasters.
In another study, which has not yet been published, after making the changes suggested by the model, non-alcoholic beer became indistinguishable from regular beer, Verstrepen said.
The idea is that brewers should aim to implement AI recommendations by tweaking recipes, rather than simply adding flavors, he says. “Simply adding pure aroma compounds is not really acceptable in beer brewing. All you need is malt, hops, water, and yeast.”
“So I look at it as a tool, specifically to be used to make better non-alcoholic beer, but not to take away the art of crafting good beer in an artisanal way,” Versträpen said. To tell.
topic:
- chemistry /
- Eating and drinking
Source: www.newscientist.com