“Wines are notoriously similar in their spectral fingerprints,” Shimazu said. “The variation this robot detects is very subtle.”
BY ERIC TALMADGE
THE ASSOCIATED PRESS
TSU, JAPAN — The ability to discern good wine from bad, name the specific brand from a tiny sip and recommend a complementary cheese would seem to be about as human a skill as there is.
In Japan, robots are doing it.
Researchers at NEC System Technologies and Mie University have designed a robot that can taste — an electromechanical sommelier able to identify dozens of different wines, cheeses and hors d’oeuvres.
“There are all kinds of robots out there doing many different things,” said Hideo Shimazu, director of the NEC System Technology Research Laboratory and a joint-leader of the robot project. “But we decided to focus on wine because that seemed like a real challenge.”
Last month, they unveiled the fruits of their two-year effort — a green-and-white prototype with eyes, a head that swivels and a mouth that lights up whenever the robot talks.
The “tasting” is done elsewhere, however.
At the end of the robot’s left arm is an infrared spectrometer. When objects are placed up against the sensor, the robot fires off a beam of infrared light. The reflected light is then analyzed in real time to determine the object’s chemical composition.
“All foods have a unique fingerprint,” Shimazu said. “The robot uses that data to identify what it is inspecting right there on the spot.”
When it has identified a wine, the robot speaks up in a childlike voice. It names the brand and adds a comment or two on the taste, such as whether it is a buttery chardonnay or a full-bodied shiraz and what kind of foods might go well on the side.
Shimazu said the robots could be “personalized,” or programmed to recognize the kinds of wines its owner prefers and recommend new varieties to fit its owner’s taste. Because it is analyzing the chemical composition of the wine or food placed before it, it can also alert its owner to possible health issues, gently warning against fatty or salty products.
Of the thousands of wines on the market, the robot can be programmed to accurately identify only a few dozen at most. It also has more trouble with the task after the bottle has been opened and the wine begins to breathe and thus transform chemically.
“Wines are notoriously similar in their spectral fingerprints,” Shimazu said. “The variation this robot detects is very subtle.”