News
business
Sports
Entertainment
Technology
Editorial/Op-ed
Travel
Gazette in Education
Advertise
Subscribe
Gazette Info
Archives
What's New
Site Map
Link to Us
Births & Deaths
Career Click
Car Click
Merchandise
Services
Miscellaneous
Real EstateRates

 

 

Thursday 7 June 2001

Software paves way for dialogue

Montreal developer designs programs allowing people to speak naturally to 'virtual agents'
AMY NEUSTEIN
Freelance


DAVE SIDAWAY, GAZETTE / Pervez Siddiqui and Mobeen Khan of Inzigo, a voice-recognition software company, are working on a platform model that might make tasks like travel arrangements easier.

Imagine ordering theatre tickets or asking for financial advice from a talking robot.

A Montreal company founded by two childhood friends - one from Montreal, the other from New Jersey - is designing novel programs that allow machines to better understand callers who are trying to book a show or get some leads on stocks.

In the shadow of Mount Royal, in the city that has become a nexus for the speech-technology industry, the inzigo laboratory - a think-tank of 23 software engineers - is designing ways to make automated customer call centres recognize the individual features of the people who access them, enabling computers to better respond to a customer's voice. In today's service world, with "virtual agents" beginning to handle everything from help-desk complaints to travel arrangements, that's a critical contribution.

"Without our proprietary and patent-pending inzigo Service Engine (iSE), programs using conventional speech recognition must predict which words a person will use to ask a question," explained Mobeen Khan, inzigo's Columbia University-educated CEO and president.

By adding a statistical element to the traditional rules-based model, inzigo's engine frees callers from following a scripted dialogue, allowing them to speak as naturally as they would with a human agent. A caller will no longer need to provide the system with information bit by bit - for instance, the city and state, followed by the name of the theatre, then the kind of show - drama, comedy, musical, but can use a normal sentence instead.

"Such natural-language applications have a much stronger appeal to customers than do prompt-based applications as they allow user interactions to be less structured, more intuitive, less frustrating and significantly more productive," Khan said.

Inzigo, whose name derives from the Dutch word inzicht (meaning "understanding"), started in Montreal two years ago with the help of some major investors in the U.S. and Finland - whose former projects include ties to Nokia, Bankers Trust and the MIT Media Lab.

Khan, described by associates as a man with a knack for bringing new technology to the marketplace, attracted these investors after successfully promoting Research in Motion's two-way pager and the Palm VII while working at Bell South Wireless Data.

Part of the appeal of inzigo's speech-recognition platform is that it is designed to dovetail with existing technology. The program will integrate with the speech-recognition engines of all the leading designers - Nuance, SpeechWorks, Philips, Locus and IBM. Inzigo's service engine will then add a crucial layer of intelligence on top of the speech-recognizers.

Inzigo hopes to have the first models of its platform up and running by the end of this year, and some leading financial institutions and event-ticketing companies in Montreal and elsewhere are already showing an interest.

The technology of speech-related computer hardware is also undergoing rapid improvement.

At the third annual Telephony Voice User Interface Conference this year in Scottsdale, Ariz., an Israeli company called Natural Speech Communications Ltd. presented an unusually "compact" system that can dramatically decrease the time it takes a caller to connect to the Voice Web - the portal system that gives direct access to the Internet via telephone, rather than through a PC. Hardware like this will give speech-based programs a vastly more efficient hookup to customers, since it is projected that within a few years one-third of all households will use voice portals to search for information online.

But even the speediest hardware won't save a system that doesn't understand what callers are saying to it. And with today's 24/7 lifestyle making it impossible for human operators to meet the demand for service, companies such as inzigo that perform sophisticated natural-language applications have an obvious market niche in the rapidly growing automated-service industry.

As the error rates in speech recognition are dropping annually by 30 per cent, the industry appears ready for inzigo's addition of intelligent understanding to voice processing.

Chester Anderson III, senior vice-president (business development) at Sound Advantage, a southern California telecom company, told the Arizona conference that the speech-recognition industry has progressed from the Information Age to the Wisdom Age.

And to think it was just 60 years ago that T.S. Eliot lamented: "Where is the wisdom we've lost in knowledge, where is the knowledge we've lost in information?"

Inzigo's goal is to make the transmission of information between humans and machines convey as much relevant knowledge as possible. That's the industry's next step on the road to wisdom.

Career Click Car Click Homes Births & Deaths Classifieds