Color, smell, taste and chemical components can be used to distinguish whiskeys
Jane Barlow/PA Images/Alamy
Artificial intelligence can distinguish Scotch whiskey from American whiskey and identify its strongest ingredients more reliably than human experts, instead of tasting the drink using data.
Andreas Grasskamp Germany’s Fraunhofer Institute for Process Engineering and Packaging IVV and his colleagues prepared an AI molecular aroma prediction algorithm called OWSum based on descriptions of different whiskeys.
Then, in a study that collected 16 samples (nine types of Scotch whiskey and seven types of bourbon or American whiskey), OWSum was tasked with distinguishing the drinks from the two nations based on key flavor descriptors such as floral, fruity, woody or smoky Using these alone, the AI could tell which country a drink matches with almost 94 percent accuracy.
As the complex aroma of these spirits is determined by the absence or presence of many chemical compounds, the researchers also provided AI with a benchmark dataset of 390 molecules commonly found in whiskies. When given AI data from gas chromatography–mass spectrometry showing which molecules were present in the sample’s spirits, OWSum increased its ability to distinguish American Scotch to 100 percent.
Compounds like menthol and citronellol were dead giveaways for American whiskey, and the presence of methyl decanoate and heptanoic acid indicated Scotch.
The researchers also tested OWSum and a neural network on their ability to predict five aroma words based on the chemical content of a whisky. With a score ranging from 1 for perfect accuracy to 0 for consistent accuracy, OWSum scored 0.72. The neural network scored 0.78 while the human whiskey expert test participants only scored 0.57.
“(The results) highlight that it’s a complicated task for humans, but it’s also a complicated task for machines, but machines are more consistent than humans,” says the team member. Satnam SinghAlso at the Fraunhofer Institute. “But that doesn’t mean humans aren’t needed: we need our machines to train them, at least right now.”
A model does not take into account the concentration of molecules, only their absence or presence, which is what the researchers hope to correct, and which could provide even greater accuracy.
Grasskamp says AI tools can be used to control quality in distilleries, or help develop new whiskeys, as well as detect fraud. But they could also be used for “anything that smells”, such as in the production of other food and drinks or in the chemical industry.
Topics:
