Whether you prefer a fruity lambic or a complex Trappist, Belgian beers have long been known for their variety, quality and heritage. Now researchers say they’ve harnessed the power of artificial intelligence to make brews even better.
Prof Kevin Verstrepen, from KU Leuven University, who led the research, said AI could help unravel the complex relationships involved in human aroma perception.
“Beer contains – like most food products – hundreds of different aroma molecules that are picked up by our tongue and nose, and our brain then integrates them into one picture. However, the compounds interact with each other, so how we perceive one also depends on the concentrations of the other,” he said.
Writing in the journal Nature Communications, Verstrepen and his colleagues report how they analyzed the chemical composition of 250 commercial Belgian beers of 22 different styles, including lagers, fruit beers, blonde beers, West Flanders beers and non-alcoholic beers.
Among the properties studied were alcohol content, pH, sugar concentration, and the presence and concentration of more than 200 different compounds involved in flavor – such as esters produced by yeasts and terpenoids from hops, both of which are involved in creating fruity notes.
A tasting panel of 16 participants tasted and rated each of the 250 beers for 50 different characteristics, such as hop flavors, sweetness and acidity – a process that took three years.
The researchers also collected 180,000 reviews of different beers from the online consumer review platform RateBeer, and found that while ratings of the brews were biased by characteristics such as price meaning they differed from the tasting panel’s ratings, the ratings and comments shared with others characteristics are related – such as e.g. as bitterness, sweetness, alcohol and malt flavor – it correlated well with that of the tasting panel.
“Small changes in the concentrations of chemicals can have a big impact, especially when multiple components start to change,” Verstrepen said, adding that one surprise was that some substances traditionally known to be a turn-off can be positive when in lower concentrations are present. and occurs in combination with other aroma compounds.
Using the different sets of data, the team built models based on machine learning – a form of AI – to predict how a beer would taste, and its appreciation, based on its composition.
They then used the results to improve an existing commercial beer, essentially enriching it with substances flagged by the models as important predictors of overall appreciation – such as lactic acid and glycerol.
The results of the tasting panel revealed that the additives improved ratings for both alcoholic and non-alcoholic beers across criteria including sweetness, body and overall appreciation.
Although the models have limitations, including that they were only developed using data sets based on high-quality commercial beer, Verstrepen said their biggest application could be in tweaking non-alcoholic beer to make it better.
But beer lovers needn’t worry that new technology could disrupt a rich heritage, with Verstrepen noting that the skill of brewers remains essential.
“The AI models predict the chemical changes that can optimize a beer, but it’s still up to brewers to make that happen from the recipe and brewing methods,” he said.
