Nestlé researchers have developed a robot that can predict the taste of coffee within minutes

Robots may soon be able to wake up and smell the coffee, research by the food giant Nestl? suggests. A team led by Christian Lindinger at the company’s research centre at Lausanne, Switzerland, has developed a machine that can correctly predict the judgements of trained human coffee tasters - and may eventually replace them. 

Coffee aficionados are able to describe the complex aroma of coffee - made up of over 1000 volatile substances - in just a few words but designing a machine that is able to do the same has proved challenging for flavour scientists. 

To establish a connection between the verbal description given by humans and the chemical analysis performed by machines, Lindinger and co-workers compiled ’sensory profiles’ for 11 different espresso coffees by asking tasters to score each one in 10 different taste categories, such as bitter, woody, or acid. 


Source: © Lindinger et al

Predicted profile (coloured patch) and human sensory profile (red line)

They then fed the results of proton transfer reaction mass spectrometry of the gas phase above each coffee into a computer program that was able to ’learn’ from experience and matched up the mass spectrometry results to the human sensory profiles. 

Finally, to test the program, the researchers used it to predict the sensory profiles for eight new types of espresso. The predicted profiles closely matched the evaluations of human tasters, they found - even though the machines were not able to ’taste’ the coffee but were restricted to analysing its smell. 

As the machine only needs two minutes to take each sniff, the Nestl? researchers say their method would be suitable for high-throughput analysis of coffee quality. From there, a machine that can brew a cup of coffee to suit every customer’s unique taste profile is surely just a small step away. 

Michael Gross