Researchers from the University of Leuven in Belgium have created an artificial intelligence model for navigating the diversity of beer. It is expected that it will be used to build a commercial tool for predicting demand for new varieties of this beverage. Beer was chosen not only because of its prevalence, but also because of the incredible complexity of its evaluation, as there are many subjective opinions that influence consumers’ liking for a particular variety.
Adverts
To train the AI, three datasets were prepared to describe in detail 250 publicly available beers in Belgium, brewed in 22 popular styles. The first type of data is scientific. The scientists measured 226 chemical parameters of the beer, mainly concentrations of various compounds. And 200 flavour parameters, also in the form of precise values.
The second set of data were evaluations from live tasters. Professional sommeliers identified 50 organoleptic parameters each and rated all the beers on them. They were most interested in flavour nuances and the sensation of impurities in the drink. They were then asked to give similar scores to enthusiastic beer drinkers who left 180,000 reviews of the selected beers via the RateBeer online platform.
