Artificial intelligence (AI) can tell the difference between American whiskey and Scotch and identify their distinct aromas, potentially better than human experts.
More than 40 different compounds with diverse structures determine the aroma profile of whisky. The chemical interactions of these compounds in the olfactory system make it extremely difficult to assess or predict the aroma characteristics of whiskey based solely on its molecular composition. Human panels are widely used to evaluate flavor, but the sense of smell can be subjective, and such evaluation methods require significant time, money, and often training. Required panelists.
So German researchers decided to see if machine learning techniques could be used to assess the molecular composition of seven American and nine Scotch whiskies. To do this, they used two machine learning algorithms. OWSum, a molecular odor prediction algorithm developed by researchers, and convolutional neural networks.
The team analyzed the whiskey samples using GC-MS and automated compound detection analysis. We also collected sensory data generated by a panel of human experts using a rating method called rate-all-that-apply (RATA) to determine the top five odor descriptors for each sample.
“Caramel-like” was the most characteristic odor descriptor for American whiskey, whereas “apple-like,” “phenolic,” and “solvent-like” odors were more pronounced in the Scotch samples. did. The researchers said this “demonstrates a clear relationship between volatile molecules, the sense of smell and the type of whisky.”
They then used the algorithm to see if it could predict the whisk’s odor as represented by the top five odor descriptors, and found that it could do so with “promising accuracy.” OWSum can determine with over 90% accuracy whether a sample is an American whiskey or a Scotch, and both algorithms average the five strongest notes of a given whisky, exceeding the average of any human expert. could be identified more accurately and consistently. .
“OWSum not only provides a quick way to classify whisky, but also allows its composition and characteristics to be analyzed in one step,” they said.
