A new artificial intelligence-based tool which includes input from scientists provides wine recommendations based on consumers’ personal tastes and budgets. The digital ‘taste map’ aims to guide drinkers towards bottles which they would naturally prefer, and claims to ‘democratise’ the process of wine selection. Rather than calling for the help of a sommelier or a wine expert, the tool aims to offer an “accessible and affordable way for individuals to discover wines”. It has been created by Danish scientists with research presented in August last year before the app was created. The information is a large multi-modal wine dataset and based on the relationship between visual perception, language and flavour. It includes 897,000 images of wine labels, 824,000 reviews of wines from the Vivino platform, and 350,000 vintages, annotated with year, region, rating, alcohol percentage, price and grape composition. Granular The scientists wrote in their introduction to the research: “Our motivation is twofold. On one hand, internet photos and user reviews are a scalable source of data, offering abundant, diverse, and easily accessible insights into wine qualities. On the other hand, human flavor annotations, while not as scalable, provide a more direct and granular understanding of the wines’ flavor profile. “By combining these resources, we aim to capture the best of both worlds, yielding a richer, more intricate dataset.” The scientists from the Technical University of Denmark, California Institute of Technology and University of Copenhagen obtained fine-grained flavour annotations on a subset by conducting a wine-tasting experiment with 256 participants who were
This Article was originally published on The Drink Business - Fine Wine
