Is it Malbec or Merlot? Ask an e-tongue

April 3, 2008 The 2-year old, 2-foot-tall, blue-and-white talking android can identify about two dozen blends and varietals by grape and region, and -- in a childlike voice -- describe them in terms like "buttery" or "full-bodied."

Electronic tongues and noses -- arrays of chemical sensors that generate "signature" patterns when exposed to specific liquids and aromas -- have been sipping and sniffing their way through our food and drink for close to 20 years, sometimes doing a better job than their more volatile human counterparts. And as the technology advances, what passes our lips is more and more likely to be rated and graded not by humans but by machines.

First unveiled in the summer of 2006 as a joint project between NEC System Technologies and Mie University in Japan, the robot sommelier, with its big eyes and bow tie and outstretched arm ready to zap a wine, is no doubt the cutest and catchiest electromechanical taster to emerge from a lab. But it's one of many gadgets around attempting to mimic the human nose and tongue.

Not that the robot actually "tastes" the wine. Instead, it blasts an infrared sensor at the wine and by analyzing the wavelengths of absorbed light, can decipher its composition and structure. Like a flesh-and-blood sommelier, it can recommend wines a customer might like by comparing the spectral "fingerprints" of different wines to the preferred wine.

It can do the same thing for various meats, cheeses and breads, and deliver what is likely unwanted health advice based on the sugar or fat composition of the foods it's testing.

"The e-nose and e-tongue work well in situations where there is an established, known taste profile of something tasty, and the food manufacturers just want to make sure it stays a certain way," says Alan Gelperin, a neuroscientist at the Monell Chemical Senses Center in Philadelphia. "You're not asking the machine to make some subtle distinction about whether it's wonderful or not wonderful. You're just asking the machine, is this the same as the one you tasted a week or a year ago?"

Research centers all across the world have scrambled to design ever-more sophisticated e-tongues or e-noses.

Concerned about its slagging tea industry, India plans to start monitoring tea quality with an e-nose developed at the Center for Development of Advanced Computing in Bangalore. Early results, according to a report in the Hindu, show a close correspondence between the machine's quality-rating scores and those of the professional tea tasters, who sip, inhale and roll the tea on their tongues.

Researchers at St. Petersburg State University in Russia built an electronic tongue that can differentiate between natural and artificial mineral waters, brands of coffee and soft drinks containing different sweeteners, and the results correlated with what a professional taste panel found.

New generations of sensors, which can consist of polymers, sound acoustic waves or, as with the robot sommelier, infrared rays, have become faster and more sensitive, said Nathan Lewis, a chemistry professor at Cal Tech. NASA's Jet Propulsion Lab built an e-nose (for which Lewis' lab developed the original sensor technology) to fly on the space shuttle and the International Space Station as a first-alert indicator of the dangerous buildup of certain gases, such as ammonia.

Other opportunities for super sniffers have emerged in disease diagnosis, homeland security, land mine, bomb, and chemical and biological warfare detection.

But of all the potential opportunities for e-tongues and e-noses, wine remains perhaps the most beguiling. "There's interesting chemistry there," neuroscientist Gelperin says. "With cake mixes the chemistry would be similarly complex, but folks wouldn't get as excited if you said you had a robot that could sample cake mixes."

Aside from the robot sommelier, a handheld Brazilian device can differentiate between cabernet sauvignons of the same vintage from two different wineries, and between different vintages from the same winery.

Andrey Legin's team at St. Petersburg State University applied one of its electronic tongues to 56 samples of Italian red wine, 20 samples of Barbera d'Asti and 36 samples of Gutturnio. The e-tongue could pick out all wine samples of the same appellation and vintage, but from different vineyards.

Australian scientists are busy studying the sensory systems of nematode worms and insect antennae to build a "new generation cybernose" for the wine industry that would indicate to winemakers the best time for picking grapes and ways to adjust the style of a wine.

And engineers continue to fine-tune the robot sommelier to get it to "distinguish more varieties of foods and wine." They're also taking steps to, well, make it more "human" by adding another sense to the robot. "A human sommelier tastes wines by using his or her total abilities, including the sense of taste, the sense of smell and the sense of texture on [the] tongue," writes Hideo Shimazu, director of NEC Japan's technology research lab and a joint leader of the winebot project, in an e-mail. "The robot lacks many senses."

But the e-tongue and e-nose technology poses some hairy challenges that can be particularly exacerbated by wine's subtle complexities. "There's more than fifty organic chemicals in wine," says Paul Keller, senior research scientist at Battelle Memorial Institute with the Pacific Northwest Laboratory, "and it also changes its odor, which we first often perceive as taste, as you're drinking it."

Not to mention that there's thousands of wine varieties and blends on the world wine market that would have to be programmed into the robot to give a full picture.

"It's not going to be the situation where you have some new vintage of wine and you want to ask the machine whether this vintage is one of the hundred best in the last forty years or something," Gelperin says. "All those subtle distinctions are such a matter of aesthetics as well as chemistry. It's going to be a long time before any machine can do these high-order things … to have all of the degrees of judgment that a wine taster would have. The machine doesn't make judgments until you teach it the judgments you want it to make, say, to recognize variability among Bordeaux."

Needless to say, the world's sommeliers need not worry for their livelihoods yet.

"The robot doesn't offend me," says Fred Dexheimer, the wine and beverage director at the BLT Restaurant Group in New York and one of 124 master sommeliers in the United States. "It's more of a gimmick than anything, but it could be a good thing maybe on the retail end and for quality assessment and authenticity."

But even if it reaches the technical heights of being able to flawlessly tell an Oregon pinot from a California pinot or a New Zealand pinot, there's still an important personal side to wine, Dexheimer says.

"The robot doesn't pick the wines for the wine list, it doesn't know the winemakers, it doesn't know the soils, it hasn't been to these places. And when people go into a restaurant, they want something recommended that someone has some experience with, and not just saying it has this fruit and that fruit.

"But the robot's fun," Dexheimer adds, "and anything that can make wine more entertaining, and anything we can do to get the word out on wine, even if it's something as crazy as this, well … "