MADRID (EFE).— Artificial intelligence (AI) never ceases to surprise with its countless applications. On this occasion it demonstrated its usefulness in predicting how consumers will rate a particular Belgian beer and what aromatic compounds can be added to improve it.
Understanding and predicting whether consumers will enjoy new food flavors is a complex task influenced by numerous chemical compounds and external factors.
The relationship between beer chemistry and people’s preferences is often investigated through consumer trials, which can be limited and make comparisons between types biased.
A team of Belgian researchers characterized more than 200 chemical properties of 250 commercial beers of 22 styles, according to a study published in “Nature Communications.”
“I wanted to have a more neutral and scientific description of the different beers in the world” to be able to compare and predict “how a beer really tastes,” explained Kevin Verstrepen, from the Catholic University of Leuven (Belgium) and one of the signatories of the research. . In addition to measuring the concentrations of hundreds of aroma compounds, each beer was evaluated on 50 criteria by a group of 15 people and included data from more than 180,000 public consumer opinions from an online beer review database, in a process that lasted five years.
Once the database was obtained, the researchers trained machine learning models to connect both variables so that the key aromas and final appreciation score of a beer might be predicted without the need for human tasting.
These results were used to improve the flavor of an existing commercial Belgian beer, to which certain aromas predicted by the model were added to increase the quality, which improved significantly in blind tastings.
The authors suggest that this tool might help improve quality control and recipe development for beers, or potentially other foods and beverages, to more effectively meet specific consumer demands.
Results are currently limited to the main commercial styles of Belgian beer, and a larger number of samples may be necessary to optimize predictions and overcome limitations, including the identification of style-specific effects and demographic information, such as age and culture.
The goal now is to make better non-alcoholic beer, says the University of Leuven.
#improves #Belgian #beer
2024-04-02 14:55:28