(This article was originally posted at http://www.nationalpost.com/search/story.html?f=/stories/20010716/613209.html.  They only archive for 60 days, so I've kept this copy here for viewing. -David)

July 16, 2001

Finding the right voice
Speech-recognition technology is rapidly becoming more reliable and convenient
Jeff Buckstein
Financial Post
 
Photo:  Matt Moyer., The Associated Press
With politicians lobbying to ban cellphones in cars, voice-recognition software in phones and hand-held computers can help drivers perform many tasks while keeping eyes on the road and hands on the wheel.

Eric Jackson remembers the primitive voice-recognition technologies that existed until about the mid-1990s. They required one to speak at length so a computer could be trained to recognize an individual's voice -- even so, there was a high rate of errors. Today, however, it is a different ballgame.

"Since then, the processors that help to process speech have increased in speed, the accuracy rates are much higher and the quality of the technology has improved," says Mr. Jackson, vice- president of marketing and business development at Toronto-based VoiceGenie Technologies Inc. "And recognition occurs in the network, so you don't have to train the speech engine to specifically recognize your voice anymore." His company specializes in building and deploying infrastructure to enable voice access to Web content.

Moreover, according to David Thomson, chief technical officer at Lucent Speech Solutions, part of Lucent Technologies Inc. in Naperville, Ill., the cost of installing speech technology in a large company today compared with five years ago may have dropped by as much as 90%.

This has provided both businesses and consumers with a number of practical applications, analysts say.

In light of the current economic slowdown, many corporate executives are prioritizing three items -- reducing costs, improving customer satisfaction and increasing revenue market share -- when it comes to e-business investment, says Ken Holl, senior manager, product management, advanced speech technologies, at Nortel Networks Corp. in Bohemia, N.Y.

Speech-recognition technology satisfies all three of those criteria, says Mr. Holl, who notes how, for instance, customer satisfaction is addressed because speech represents a more natural way of communicating. "People would much rather speak directly to somebody versus speaking to a machine. It provides a more natural, intuitive interface," he says.

Voice-recognition software encompasses different types of applications. One, for instance, involves a voice portal, providing callers with access to information such as traffic, weather and news reports and stock quotes. Through their phone, users will be able to get the type of information they now receive from their personal computer using a browser.

Another service involves the ability to perform tasks such as place or screen calls, take a message, schedule a meeting or check calendars. This provides users with several advantages. If, say, business travellers are at the airport without their computer and wants to know if a meeting has been cancelled, they can simply use voice commands to call and check their voice mail, then have it read to them over the telephone, says Mr. Thomson.

Also, notes Mr. Jackson, many jurisdictions in Canada and the United States are beginning to limit the use of hand-held cellphones when driving. As a result, North American telecommunications companies are examining how to offer voice-activated services to drivers, so they can keep their full attention on the road.

One business area where voice- recognition technology has proven to be extremely popular, offering large cost savings, is in call-centre applications.

"Voice communication has been most popular for companies who want to off-load the number of calls going into live agents. It's very expensive to build the whole call-centre infrastructure. So even if the company only off-loads a small number of calls, they end up saving a huge amount of money," Mr. Jackson says.

Another major development in the field of voice communications is VoiceXML, which provides a standard language for voice access to Web content. Lucent, AT&T and Motorola Inc. formally announced they had teamed up to develop that service in March, 1999; IBM subsequently joined them the following month. Representatives from the participating companies felt that the technology was so important, it was vital to "form a standards body to create a new [common] language," Mr. Thomson says.

Prior to the development of this standard language -- a first version was running by August, 1999 -- there had been several proprietary speech-recognition applications; however, they were incapable of communicating with each other.

The development of a standard language has also contributed positively toward cost savings. "If you're building a small service, development costs dominate. But VoiceXML and similar technologies are now making it possible for the average guy to just go in and build a service very quickly," Mr. Thomson says.

Another overall benefit is that developments in this field have caught the attention of the markets, despite the current general slowdown, says Peter Doyle, Toronto-based national industry leader, software and electronics, at KPMG.

He notes that even with last year's meltdown in many IPO and capital-funding markets, voice-recognition technology has continued to attract a lot of venture-capital financing. "This tells me it's a market that everyone believes still has a lot of opportunity," Mr. Doyle says.