Stellar Startup: Training your PC, your way

I used to have this talent, back in the day: If a computer was misbehaving, all I had to do was look at it, and it would start or stop doing whatever it was supposed to do, or not do.

LINGUISTIC AGENTS 88 298 (photo credit: )
(photo credit: )
I used to have this talent, back in the day: If a computer was misbehaving, all I had to do was look at it, and it would start or stop doing whatever it was supposed to do, or not do. True story. For years, I was able to "communicate" with computers, apparently using some subconscious mind-meld technique, to get them to do what I wanted. Once in awhile, I'd come across a truly obstinate case, and I would have to give it a verbal drubbing, depending on the situation. Sometimes a hint was sufficient, other times an outright threat was the only way. In the end, they did what I wanted them to do! Unfortunately, only a few of us are blessed with such extra-sensory talents. Most people who interact with computer technology are forced to slog through complicated instruction sets - in the form of line commands, menus or even, in the case of many new Web sites geared for the cell phone market, verbal instructions. In each case, you have to be very specific in indicating what it is you want the computer to do - or risk getting the equivalent of a "dumb stare," the type kids are famous for when they don't feel like doing their chores. It's like you can hear the computer saying "Huh?" Now, we know that the kids understand exactly what we want when we ask them to do an errand. They're smart, as their report cards indicate (and heaven help them if those cards are sub-par!). So we're not fooled by the "duh" look on their faces when it comes time to work. And yet, we're willing to accept that behavior from computers, which are obviously pretty smart creatures. But unless you have a natural talent for intimidating computers, you're pretty much going to have a frustrating experience getting a computer to fulfill your request - whether it's scrolling through a dozen menus or submenus to find the right command, coming up with the right text combination when doing a Web search, or finding the "magic word" that will prompt the system to fulfill your verbal request. But never fear. If the folks at Linguistic Agents have their way, you too will be able to easily and quickly tame your computer, making it do the work of understanding what you want and fulfilling your desires - instead of putting the onus of proper communication on you. As human beings know, meaning can be difficult enough to gauge between two intelligent people with an understanding of the context of communication. Computers may have a lot of "book smarts," says Simcha Margaliot, chief operating officer of Linguistic Agents, a Jerusalem company that intends to teach PCs, databases, Web sites and cell phone services what people mean when they speak. Imagine, he says, calling up an information service that allows users to get weather forecasts by asking for them verbally, as opposed to the "phone menu" method in use by most services today, where subscribers press a series of buttons on their phone to access the desired information. Such services, like this one (, have been around for several years now. But the problem with most of them is that they understand a "limited vocabulary," to say the least. Try, for example, asking any voice information service the date of the next "March March March" ( and see what answer you get! It's all about context, says Margaliot. A "Huh?" reaction by the database to the super-homonym querying the date of the annual parade by the residents of March, England, on the last Saturday of the third month of the year is just one example of how much work still needs to be done in order for us to be able to communicate with computers using "natural language," which is basically the lingua franca in use by the folks dialing the service. Computer communication context confusion isn't just an issue around the Ides of March; simple tasks like presenting a weather forecast to users would be much more useful, and practical, if Linguistic Agents had its way, Margaliot said. For example, try this Google search: "New York City" + "Do I need a coat today." Google will give you its equivalent of a teenager's "Huh?," answering "your search did not match any documents." And while it's easy to type "New York City Weather" into a search box on your PC, it's a mighty task if you're using your cell phone to connect to WAP weather sites. Getting databases to understand natural language is going to be a major issue in the coming years, as cell phones become the primary data tools for billions of people around the world. "The only practical way to poll for information on a cell phone is via text messaging. While we're willing to go through dozens of search results on our 17-inch computer screen, small screens on mobile devices mean users want to send a query and receive an accurate message. But you have to first understand the query. This is why there is a much greater reliance on natural language with mobile phones," Margaliot says. Which is where Linguistics Agents ( comes in. Margaliot says the company has the most advanced natural language engine in the business. Most of the other companies working in this space, like Nuance (, deal more with grammar and vocabulary, essentially expanding the limits of human-database communication. But you can only go so far down that road, he says. Instead, Linguistic Agents has built its technology on the latest human linguistics scene and theory, most specifically Nanosyntax (, which basically believes that words can be broken down into "subatomic particles," the better to be manipulated and rebuilt via natural language programming into a communications interface resembling human language as closely as possible. Although Nanosyntax theory has been around for a while, as have natural language programming projects, no other company has come as far as LA, Margaliot says - because the company's developers are not just programmers but linguists, as well, as is company's founder and chief executive officer Sasson Margaliot. And, says Simcha Margaliot, you need a linguist's touch when dealing with what is essentially translating linguistic theory into practical programming. As a company, Linguistics Agents has been around for about six years, but Margaliot says that LA still qualifies as a start-up - because the linguistic theories the company is basing its technology on has changed radically over the past few years. "There has been more advancement in natural language sciences since the year 2000 than there had been over the previous two decades, and Linguistics Agents has been at the forefront of that development," says Margaliot. For example, a practical application based on LA's development is nearing completion, and close to the hearts of the company's hareidi ownership: InfoJew ( will allow observant travelers to Europe access to information on services, kosher restaurants, databases etc. from their cell phones using natural language queries. If that works out, and Margaliot is sure that it will, we'll soon be able to ask our computers or cell phones any question we want, the way we want, and get the answers we\re looking for. Our kids might still say "Huh?" when we ask them to do stuff - but at least the databases we query will do what we tell them to!