Researchers from Denmark and the United States have trained an algorithm to distinguish between the flavors of different wines.
Apps such as Vivino and Hello Vino operate on algorithms enabling prospective buyers to scan wine labels and gather information about the contents of bottles and read comments from other buyers.
Now, researchers from the Technical University of Denmark (DTU), the University of Copenhagen, and the California Institute of Technology (Caltech) have incorporated another highly significant parameter into algorithms, making it much easier to find a wine that aligns with our taste buds! They fed an algorithm with the taste impressions left by different wines on different individuals, as per a study published on the open-access platform arXiv.
“We demonstrated that by feeding the algorithm data related to the taste impressions of different people, it was able to make more accurate predictions about the type of wine someone would prefer,” stated Thoranna Bender, a graduate from DTU and the lead author of the study.
Researchers engaged 256 participants in a wine-tasting process. Volunteers placed cups with different wines on an A3 paper collage based on which wines they thought tasted similar. The greater the distance the volunteers placed between the glasses, the greater the perceived taste difference they believed existed. Subsequently, scientists digitized the points on the paper collages by photographing them.
In the next step they combined the data obtained from digitization with hundreds of thousands of wine labels and user reviews provided by Vivino. Based on this massive dataset, the algorithm of taste was developed.
“This model provides information on which wines have similar tastes and which do not. So, for instance, one can hold their favorite bottle of wine and, through the algorithm, find out which other wine is similar both in taste and price,” explained Thoranna Bender.
Thoranna Bender emphasized that this new method can be applied to other types of beverages and food as well. ” If we manage to better understand the similarities in taste among different foods, we could use the method in the healthcare field to create dishes that cater to taste preferences as well as the dietary needs of patients.”
