The authors recorded the multidimensional response generated by the array of sensors and how this was influenced by the type of beer considered. An initial analysis enabled them to change coordinates to view the grouping better, although it was not effective for classifying the beers.

“Using more powerful tools — supervised learning — and linear discriminant analysis did enable us to distinguish between the main categories of beer we studied: Schwarzbier, lager, double malt, Pilsen, Alsatian and low-alcohol,” Del Valle continues, “and with a success rate of 81.9%.”

Furthermore, it is worth noting that varieties of beers that the tongue is not trained to recognise, such as beer/soft drink mixes or foreign makes, were not identified (discrepant samples), which, according to the experts, validates the system as it does not recognise brands for which it was not trained.