Color, odor, taste, and chemical composition can all be used to differentiate whisky.
Jane Barlow/PA Images/Alamy
Using data instead of tasting, artificial intelligence can distinguish between Scotch and American whisky, accurately identifying the aroma of its primary ingredient better than a human expert.
andreas glaskamp Researchers from Germany’s Fraunhofer Institute for Process Engineering and Packaging IVV trained an AI molecular odor prediction algorithm called OWSum based on descriptions of different whiskies.
In a study with 16 samples of nine Scotch whiskies and seven American bourbons or whiskies, OWSum, based on keyword descriptions like floral, fruity, woody, or flavored, was able to differentiate between the two countries with nearly 94% accuracy. By also providing a dataset of 390 molecules commonly found in whisky, OWSum’s accuracy in differentiating American and Scotch dram increased by 100% when given data from gas chromatography-mass spectrometry.
Compounds like menthol and citronellol distinguished American whisky, while methyl decanoate and heptanoic acid were characteristic of Scotch.
The ability of OWSum and the neural network to predict the top five odor keywords based on whiskey’s chemical composition was assessed. OWSum scored 0.72, the neural network scored 0.78, and human participants achieved 0.57.
“[The results] highlight the complexity of the task for humans and machines. Machines, however, are more consistent,” said a team member, Satnam Singh, at the Fraunhofer Institute.
Both models do not consider molecule concentration, only presence or absence, which researchers aim to improve for higher accuracy.
Glaskamp suggested using AI tools for quality control in distilleries, developing new whiskies, and detecting fraudulent products in various industries.
topic:
Source: www.newscientist.com
