Posted: 02/2000

Speech Recognition Blasts into Future
By Liz Montalbano

Speech recognition technology took the giant step in 1999 from testing to deployment in
such areas as directories, information retrieval and automated attendants. While industry
experts say 2000 will be a year to hone those advances, others say the technology will
spawn the deployability of far more complex applications.

These include enabling people to use their telephones as personalized web browsers, or
allowing business auto attendants to identify incoming callers by stored voiceprints.

This year also may see the nationwide deployment of speech recognition-enhanced
directory assistance (see "The Name Game: Speech Recognition Ushers DA into
2000," PHONE+, December).

"What we’ve seen over the course of this last year is actual deployment of
production systems–it makes for a significant difference from prior years, where people
were playing with it and toying with it," says Brian Spraetz, product manager for
speech products for InterVoice-Brite Inc. (www.intervoice.com).
His company integrates speech recognition technology into its interactive voice response
(IVR) platforms. "Now it’s actually out there working for customers."

Spraetz cites speech recognition-enhanced auto attendants in the airline industry as an
example of how companies can deploy the technology, and replace various job functions
previously filled by people.

"[Airlines are] not just using it for flight information, they’re also using it
for reservation information, and for baggage tracking," Spraetz says.

He notes the latter use of the technology "took the heat off live agents,"
because "people tend not to yell at a computer" if they’re trying to locate a
lost bag.

The Players

InterVoice-Brite actually doesn’t develop speech-recognition technology. The company
integrates it into deployable platforms to sell directly to customers and
resellers–including service providers such as Williams Communications (www.williams.com). Williams then integrates the
platforms with its voice products and sells them to end users or network providers such as
Siemens AG (www.siemens.com), which integrates the
platforms into communications network solutions.

InterVoice couldn’t achieve such integration without strategic partnerships with actual
speech recognition technology providers, such as Nuance Communications (www.nuancecom.com) and SpeechWorks Intern-ational Inc.
(www.speechworks.com).

Nuance Marketing Vice President Steve Ehrlich says his company has made significant
inroads in developing voice authentication systems. These systems use "biometric
identification" to make a voiceprint of someone’s voice–much like a fingerprint. The
voiceprint can be used in automated systems, which could lead to the elimination of
customer personal identification numbers (PINs).

"What a voiceprint does is when the user calls up, instead of having to say the
PIN number, all they have to do is say their account number or their name," Ehrlich
says. "Then the system checks their voice against a voiceprint that’s enrolled in the
database."

The Home Shopping Network uses the first large-scale deployment of the technology,
Ehrlich says. With the network’s success using the technology, he expects other companies
to follow, as they and their customers become more comfortable with it.

Another advancement representing the future of speech recognition technology comes in
the form of Nuance’s Voyager, Ehrlich says. This voice browser permits people to place
phone calls and reach voice sites on a voice web.

"What Voyager allows you to do is provide a standard user interface to
applications and navigate between different content sites," Ehrlich explains.
"It would be offered as a service to customers either by a carrier or some other
service provider."

Ehrlich says a voice prompt would greet users when they called into Voyager with
"How may I help you?" From that point, users could tell the system they would
like to place a phone call or purchase a plane ticket via e-commerce, depending on what
services the "voice portal" is programmed to do.

"Voyager has a user interface very much like a Netscape [Navigator] or [Internet]
Explorer on the desktop, except it’s on the telephone," Ehrlich adds. "When you
speak to Voyager, you can say, ‘go to my bookmarks,’ or, ‘go to this site’ and when you’re
done, you can say, ‘go back.’"

Nuance also uses its speech recognition technology to improve customer service
telephone systems. At press time, the company announced it is working with Spanlink
Communications Inc. (www.spanlink.com) and US WEST
Inc. (www.uswest.com) to develop a system that would
enable US WEST customers to make service requests via a speech-recognition enhanced IVR
system rather than using a touchtone system or wait for a customer service representative
(CSR).

With the technology, a customer could state the problem, such as "My phone line is
out," rather than using a touchtone system or speaking to a live agent. Callers also
can use the technology to make requests for services including repairs, adds, moves,
billing information and customized voice and data services.

SpeechWorks’ marketing vice president Steve Chambers says his company is focused on
telephony-based speech applications as well. He cites three categories on which
SpeechWorks currently is working: speech for communications, such as auto attendants and
company directories; information retrieval, such as applications that will allow customers
to call into companies and obtain product information; and transaction applications, which
will allow end users to use speech recognition to negotiate revenue transactions over the
phone without talking to a live agent.

He says his company is developing tools for partners to build speech applications, and
"we also have packaged applications as well."

"SpeechSite, for example, is a packaged auto attendant and information retrieval
application companies can buy," he explains.

As an example, he cites a SpeechWorks’ product that companies, such as pharmaceutical
supplier McKessonHBOC Inc. (www.mckhboc.com), use
that enables customers to place or check orders and to access other customer service
features. This is all done by using speech recognition technology.

SpeechWorks also has launched or tested significant deployments of systems that allow
international companies to offer stock quotes over the phone with speech
recog-nition-enhanced applications.

In Australia, SpeechWorks teamed with TD Waterhouse Australia, a division of TD
Waterhouse (www.tdwaterhouse.com), to deploy
TalkBroker. This product allows clients to get stock quotes and other trading information
on the phone when they don’t have access to a computer.

A similar system in France allows the public to access free stock quotes during a
month-long period. The available data are from The Paris Bourse Regelement Mensuel, the
French stock exchange, in a co-development deal with GL Multimedi@, a division of GL Trade
Group (www.gltrade.fr).

The Root of the Matter

Mark Bannon, vice president of sales, marketing and technical services for Phonetic
Systems Inc. (www.phoneticsystems.com),
another speech recognition technology provider, is more interested in some of the
applications driving this advanced technology.

When the voice portals topic was broached, Bannon says bluntly, "No one knows what
the hell they are." He then admits that just because the technology isn’t there yet,
it doesn’t mean the industry isn’t buzzing about it.

"That technology is being developed, but it’s far off. [But] we just went out and
got millions of dollars worth of venture capital money, and one of the reasons people gave
it to us is that we’re going to develop a speech-enabled Internet portal," Bannon
says.

Bannon adds the root of technology, such as the voice portal, is "personal
information systems."

"People put personal information systems on the Internet, then people can get at
them from anywhere," he says. "Some of that might be around next year, but I
think this industry has to get more realistic about the applications rather than the
technology."

To Bannon, it’s more realistic to believe some applications deploy more easily than
others. An example is speech recognition-enhanced business directory services, which would
allow callers to perform advanced search options to reach the person they want.

For example, iVoice.com Inc. (www.ivoice.com),
which manufactures voice and computer telephony systems, recently integrated a speech
recognition-enhanced directory function into its existing 2001 Voice Mail system. To
develop a speech recognition-enhanced auto attendant for small to medium-sized businesses,
iVoice.com integrated the technology of Entropic Inc. (www.entropic.com),
a speech recognition technology developer, which Microsoft Corp. (www.microsoft.com) recently acquired.

Bannon also says he believes the speech recognition technology’s future is rooted in
directories for the small office/home office (SOHO) market, which will allow those firms
to link employees in multiple locations to a virtual receptionist.

"So I call up and it says, ‘Welcome to ABC Company; please say the name,’ and you
get transferred to that extension, even though one guy might be in California, one guy
might be in Chicago etc.," Bannon explains. "It’s the concept of really
centralizing the auto attendant capabilities for a virtual company, no matter what size
company it is."

He adds that such directories will be a value-added service for carriers to offer
business customers.

"If you say to someone, how about for 50 bucks a month or $600 a year, this auto
attendant will basically act as a much more professional receptionist to you," Bannon
says. "Now times that by 50,000 customers at 50 bucks a month, and all of a sudden
they have a very nice revenue stream, and it’s a nice service for the client and for the
client’s customers."

The Payoff

Since boosting the bottom line basically is what drives service providers or companies
that may consider using speech recognition technology, it’s important the technology
proves it can pay for itself, Spraetz says.

According to his company’s research, it does.

"Typical paybacks that we’ve seen have gone from six months to a year,"
Spraetz says. "But the main savings come through saving agent time and that can
either be used to letting those agents concentrate on other tasks or to help with all
center growth."

Ehrlich agrees that using speech recognition to replace live operators in call-center
environments saves money. He adds Nuance research finds return on investment (ROI) happens
in three months, not six.

Ehrlich also cites the implementation of speech recognition technology in call centers
of Sears, Roebuck and Co. (www.sears.com) as an example
of repositioning call-center agents for maximum efficiency.

"They had 3,000 agents spread around 800 stores in the United States, and all they
were doing was answering the phone and transferring a call to a department," he says.
"They took 2,800 of those 3,000 people and put them in some other job."

According to a published Nuance white paper on the cost benefits speech recognition
provides, companies can save 93 cents per call using the technology over people.

That adds up to more than $1.5 million in a year and an 86-day payback period,
according to the company (see chart, "Payback Analysis for Speech Recognition").

Graph-Payback Analysis
for Speech Recognition

Nuance also compared the per-call costs of live agents versus speech-enabled IVR
systems. It figured the former costs $1.02, while the latter costs merely 9 cents (see
chart, "Key Comparisons Between Agents and IVR," below).

Graph-Key Comparisons
Between Agents and IVR

While SpeechWorks’ Chambers agrees that speech recognition applications can save money,
he concedes that not everyone is thrilled at the idea of implementing them.

"I’ll be honest with you–it’s new and it’s one of those things where people say
‘Prove it to me,’" he says. "People don’t feel the systems work as well as they
do, and that is very, very heavy out there in terms of perception."

He says this is why in marketing speech recognition to potential customers,
Speech-Works uses large-company deployments as examples. Among those would be the
automated voice systems of United Airlines Inc. (www.ual.com) and Federal Express Corp.
(www.fedex.com).

"We do that to legitimize the technology," Chambers says. "Then we’re
going to do a lot of experiential marketing–a lot of things that make people interact
with the system and start to suspend that disbelief."

Liz Montalbano is news editor for
PHONE+ magazine.