Neural network will help Belgian brewers brew beer that everyone will like
Scientists from the University of Leuven have created an unusual neural network that can take beer production to a new level. With its help, experts expect to achieve the perfect taste of a foamy drink based on machine learning on a large sample of experts and consumers. This was reported by Nature.
Artificial intelligence was trained on data from hundreds of experts and thousands of ordinary consumers. To generate a complete set of beer flavor data, the researchers selected 250 Belgian beers brewed using 22 different technologies. Then 226 chemical parameters of each type of beverage were identified, including acidity, alcohol, sugar, and flavoring additives.
We also formed a special tasting committee that evaluated each variety based on 50 sensory attributes, mostly taste. To supplement the opinion of the commission’s experts, the scientists processed another 180,000 reviews of the varieties under study from ordinary consumers.
The result was a large data set with numerical estimates of the appearance, aroma, taste, richness, and other characteristics of beer. Consumer feedback, commission scores, and chemical parameters were combined to train artificial intelligence. The neural network had to identify the parameters and properties of the ideal drink that would suit consumers the most. She noted the ethyl acetate content as the most important parameter that affects the evaluation of beer. It usually conveys the nuances of flavor: fruity, malt, or alcoholic. Ethanol is noted as the second most important component. It affects both the taste and physical parameters of the drink. The status of lactic acid is also recognized as significant.
The resulting model allowed us to predict various characteristics of beer and its appeal to certain consumer groups. An experiment was also conducted with the components of the drink in a non-alcoholic version. Experts noted a significant improvement in the taste of the foam.
The authors of the study expect that the neural network will improve control over the quality of beer products and help create new tasty varieties. The model can be improved by adding different types of beverages from other regions and prepared using other technologies.
