Researchers at KU Leuven in Belgium have developed a novel machine learning approach aimed at creating new alcoholic and non-alcoholic Belgian beer flavors to enhance consumer satisfaction and streamline production processes. Their study, published in the journal Nature Communications, led by Kevin Verstrepen and his team, meticulously analyzed over 200 chemical properties of 250 commercially available Belgian beers, categorized into 22 types including Blond and Tripel. By correlating this data with sensory profiles collected from a trained panel of 16 tasters and reviews from the RateBeer database, which houses over 180,000 beer critiques, they trained an artificial intelligence model to predict taste and consumer appreciation levels based on each beer’s chemical composition.

This innovative AI model showed promising results in modifying both alcoholic and non-alcoholic beers, leading to higher overall consumer appreciation, as validated by expert tasters. The implications extend beyond the realm of beer production; the researchers suggest that this tool could be leveraged to enhance quality control and recipe development in other industries, catering to specific consumer preferences.

While the study’s findings are currently limited to Belgian beer alternatives, the researchers acknowledge the potential for broader applications and plan to expand their dataset to optimize predictions. They also highlight the importance of considering various demographic, personal, and cultural factors that influence consumer preferences in future research endeavors.

Lascia un commento

Il tuo indirizzo email non sarà pubblicato. I campi obbligatori sono contrassegnati *