Choosing a bottle of wine can be daunting for those unfamiliar with the plethora of labels on the shelf. Wondering about the taste and recalling past favorites can be a challenge. Luckily, apps like Vivino, Hello Vino, and Wine Searcher come to the rescue. These tools allow users to scan labels, access wine information, and read reviews, all thanks to advanced AI algorithms.
Now, researchers from the Technical University of Denmark, the University of Copenhagen, and Caltech have enhanced these algorithms. They incorporated a unique parameter – people’s flavor impressions. By feeding the algorithm with data from individuals’ taste preferences, they achieved more accurate predictions of personal wine preferences.
During wine tastings with 256 participants, the researchers had them arrange shot-sized cups of various wines based on perceived similarities. This data, along with information from hundreds of thousands of wine labels and user reviews from Vivino, formed the basis for the new algorithm. The combination of human sensory experiences with traditional data sources improved the accuracy of predicting people’s wine preferences.
This innovative approach is not limited to wine. According to Professor Serge Belongie from the University of Copenhagen, the method can be extended to beer and coffee. The use of multimodal data, including taste and sensory inputs, holds great potential in the food sector. This groundbreaking research demonstrates the power of integrating human-based inputs into artificial intelligence, opening avenues for further exploration at the intersection of food science and AI.
Source NeuroScienceNews
