‘We thought to ourselves – if we could apply the latest breakthroughs in modern linguistics to software, the results would be amazing’ – Linguistic Agents’ CEO Sasson Margaliot. It’s been called the ‘holy grail’ of computer technology – software that …
The company’s Advanced Language Machine – or ALM – is the brainchild of Linguistic Agents’ founder and CEO Sasson Margaliot. The 47-year-old Margaliot, who studied linguistics at UCLA and computer science at Hebrew University, spent years developing software for other companies before putting his attention towards creating an engine that could accurately decode human language.
“Our team was frustrated by the fact that even the most advanced software in existence was using old tools such as statistical analysis and outdated linguistic theories,” Margaliot told ISRAEL21c. “We thought to ourselves – if we could apply the latest breakthroughs in modern linguistics to this field, the results would be amazing. It would be like a placing a pot of gold on the table for some software companies!”
The ALM diagrams sentences using a recent advance in theoretical linguistics known as NanoSyntax. NanoSyntax, developed by linguists at the Massachusetts Institute of Technology in 2004, breaks language down into its component parts, going beyond words to parse the meaning of sentences. It was the introduction of NanoSyntax, with its greater ability to decode ambiguity, which, according to Linguistic Agents, has made their innovation possible.
“Most companies in the field are working with computational linguistics which is a different field, less advanced [than NanoSyntax],” explains Sruly Taber, Linguistic Agents’ Chief Information Officer.
As he describes it, computational linguistics works to understand bits of meaning from individual words and what you can extract from them, rather than trying to decode meaning in a phrase or sentence. In contrast, he continues, “NanoSyntax is saying we can understand sentences.”
To put it another way, NanoSyntax describes how the brain takes the information embedded in language and understands it. This, according to Taber, is far from simple.
“The brain has an ability to understand language without knowing the words,” he explains. “If say a sentence with a word you may have never heard, through the context, you’d be able to understand 70% of what I’m saying without that word, or even understand what that word means, just by the context. That’s what makes us very special.”
Using the core technology of NanoSyntax, Linguistic Agents’ computer scientists and programmers set out to create a technology that would mirror this process – only backwards. Instead of taking meaning from the brain, where it’s created, Linguistic Agents needed to figure out a way to take sentences and convert the information in them to a format, a linguistic tree, that would enable a computer to understand their meanings. It has taken six years of development, but now Margaliot and his staff of eight have alpha and beta versions of the technology.
“To finally have a working product is a dream come true, quite literally,” Margaliot says. “Our company is the first to enable computers to ‘understand’ natural language at this level. Knowing this means we feel not only self-accomplishment but also responsibility for making the most of this.”
The company has successfully adapted its software, known as the Intelligent Command Engine, for use in a bus system and a yellow pages directory, although these are not yet being used commercially.
In a recent demo, Linguistic Agents’ staff showed ISRAEL21c what a search in their new yellow pages directory would look like. Given the search phrase “I’m looking for a hotel,” the directory came back immediately with the question, “where?”
“That’s way beyond a search engine,” claims Taber, “because Google, even if it would somewhat be able to understand (which it won’t), would just look for the words ‘look for’ and ‘hotel’.”
Instead, the Intelligent Command Engine understands the question, and that it needs more information to provide the answer. And the engine retains the information it has already received, so that when we responded “Jerusalem” to its query, it was enough for the program to respond with a list of hotels in the city.
The possible applications seem nearly limitless. By being able to ask a computer, in natural language, what a person is looking for, search engines could be improved, automated help lines could be more efficient, information could be more easily accessible on websites – the list goes on and on.
For now, Linguistic Agents’ is focusing its marketing on companies with customer relations needs, primarily through the web and SMS. “Anybody who has a body of information that they want people to ask questions about – they can integrate it with this technology in a very short period of time,” Taber says.
And, as the world quickly shifts to handheld devices with multiple functions, Linguistic Agents’ Intelligent Command Engine will, they claim, become even more valuable by helping make these intelligent devices more user-friendly.
“Just think about how complex your little PDA (Personal Digital Assistant) is going to be when it has your phone, your computer, your TV, your radio and maybe even your microwave in there,” offers Taber. “When you think about it… it will take an hour to figure out who you’re meeting tomorrow, because you’ll have to get past your TV in there. Unless you can say – who am I meeting tomorrow? Or type it in, and it will get back to you.”
Currently, the software is available only for written, or ‘typed in’ language, but Linguistic Agents is hoping to have voice-enabled technology ready by next year.
One of the main advantages of the software is that it operates on minimal system resources and is therefore easily integrated into existing applications. And, while the beta version exists in both English and Hebrew, because of the sophistication of the NanoSyntax model, it is easily adaptable to most languages with only small adjustments. Linguistic Agents’ close-knit team has high hopes for its technology.
“I think bringing computers to a new generation is so big… it’s hard to say how big it will be,” says Simcha Margaliot, the company’s Senior Vice President and Chief Operating Officer. “This engine can go in almost every little kind of machine, computer, application. If you’re looking for information, if you want the machine to do something, you always need to get the computer to understand what you want it to do. And what we are doing is making it a lot easier – the same way you talk with friends, you’ll talk to your computer.”
“In the last 10 years, computers have taught people how to talk to them,” continues CIO Taber. “It’s sort of ludicrous. We built these things, and then we had them train us how to talk to them. Type this, put a quotation mark, or a minus, a plus, this or that – we have to talk to them in codes. Shouldn’t we be able to talk the way we do? If we do that, it’s a revolution, an entire way anything computes, and everything computes.”