AI detects whiskey aromas better than human sommeliers

Machine learning algorithms were able to identify the predominant aromas of different whiskies

According to a study published this Thursday (19), machine learning algorithms were able to identify the aromas predominant of various whiskies more effectively than a specialist.

Most odors around us is made up of a combination complexes that interact with our olfactory system to produce a particular impression. This is the case of whisky, which has an aromatic profile defined by more than 40 compounds, as well as being able to include an even greater quantity of non-odorous volatile compounds.

This makes it particularly difficult to evaluate or identify the aromatic properties of a whiskey based solely on its molecular structure. However, chemists were able to achieve this feat thanks to two machine learning algorithms, as shown by the results of research published in Communications Chemistry.

OWSum, the first algorithm, is a statistical tool that identifies molecular odors, created by the study authors. CNN, on the other hand, is a convolutional neural network that helps identify relationships in extremely complex data sets, such as those involving “the most impactful molecules and flavor attributes” in a blended whiskey, as Andreas Grasskamp explained, ​​researcher at the Fraunhofer Institute for Process Engineering and Packaging IVV in Freising (Germany), and lead author of the study.

The scientists “trained” the algorithms by providing a list of molecules identified by gas chromatography and mass spectrometry in 16 whiskey samples, including Talisker Isle of Skye Malt (10 Year Old), Glenmorangie Original, Four Roses Single Barrel, Johnnie Walker Red Label and Jack Daniel’s, among others.

They also provided specific flavor descriptions for each sample, determined by a panel of 11 experts. Next, algorithms were used to determine each whiskey’s country of origin and its top five notes.

OWSum was able to accurately identify whether a whiskey was American or Scottish. The identification of compounds such as menthol and citronellol was strongly linked to the categorization as American whiskey, while the identification of methyl decanoate and heptanoic acid was more linked to Scotch whiskies.

In a second step, the scientists asked OWSum and CNN to predict the olfactory properties of whiskeys based on the identified molecules or their structural properties.

The two algorithms were able to identify the five predominant notes of a whiskey more accurately and consistently than any human expert on the panel. “We found that our algorithms aligned better with the panel’s results than with each individual member, thus providing a better estimate of overall odor perception,” said Grasskamp, ​​one of the researchers.

These machine learning methodologies could be used to spot fakes or to evaluate whether blended whiskey “will have the expected flavor, thus helping to reduce costs by limiting the need for evaluation panels,” he says.

Source: Terra

Rose

Rose James is a Gossipify movie and series reviewer known for her in-depth analysis and unique perspective on the latest releases. With a background in film studies, she provides engaging and informative reviews, and keeps readers up to date with industry trends and emerging talents.