A group of scientists has created an artificial intelligence tool by teaching it an algorithm to track the origin of wine based on chemical analysis. The Guardian writes about it, UNN reports.
Details
According to a new study, published in Communications Chemistry, scientists have managed to develop an AI tool to distinguish wines based on subtle differences in the concentrations of dozens of compounds - this, as indicated, will allow to track wines in practice not only in a specific viticultural region, but also in the estate where they were produced.
There's a lot of wine fraud around: people make up some crap in their garages, print labels and sell them for thousands of dollars. ... We show for the first time that we have enough sensitivity with our chemical methods to tell the difference
To develop the
program, the scientists turned to gas chromatography, which was used to analyze 80 wines collected over a 12-year period.
to analyze 80 wines collected over 12 years from seven different estates in the
Bordeaux region in France. This technique is commonly used in laboratories to
to separate and identify the compounds that make up a mixture.
SEE ALSO: The first 30 Ukrainian teams presented innovations for humanitarian demining
Instead of trying to find individual compounds that distinguish one wine from another, the algorithm relies on all the chemicals found in the wine to produce the most reliable signature for each of them. The program displays its results on a two-dimensional grid, where wines with similar signatures are grouped together.
The first thing we saw were clusters that corresponded to certain chateaux (wines belonging to the highest category in the French wine classification - ed.) This immediately told us that there is a specific chemical signature for each chateau, regardless of vintage
According to the scientist, there is a general pattern of concentration of many molecules that distinguishes châteaux, but "each of them is a symphony: there is not a single note that distinguishes them, it is a whole melody."
Supplement
The study published in Communications Chemistry suggests that machine learning can help in fraud investigations by confirming whether a wine is true to its label. In Europe, where counterfeit alcohol results in an annual loss of sales worth 3,000 million euros annually, recent cases have highlighted the extent of of criminal networks operating in this industry. Earlier this year, gang members were convicted of smuggling tankers of Spanish table wine into France and then passing it off as French. The fraud had been going on for years and is believed to have involved believed to have involved the equivalent of nearly 5 million bottles, some of which were repackaged as Bordeaux.
While the detection of the fraud is the most obvious application of the program, according to scientists, this approach can also be used to monitor quality throughout the entire the winemaking process and ensure proper blending.
SEE ALSO: No-selfie zones set up on Goa beaches