By C.J. Kennedy
February 02, 2001
Level Two: Voice Browsing the World
Wide Web
The next level of technology is voice browsing websites that
offer voice portals. In attempt to maintain customer loyalty,
portals like myYahoo are offering voice browsing of personal
content. AOL recently bought Quack.com in an attempt to claim
this market too. Look for Excite and Netscape to become involved
as they can not afford to lose customers to churn.
Europe has taken a more aggressive approach to the voice
portal market. The United States, with over 40% of homes connecting
to the web by PC's, has no immediate demand for web connection
through phone. However, in Europe PC penetration rates are
much lower and people are purchasing internet-ready phones
at a rate of 5 to 1. Italy has only twenty percent PC penetration
rates but over fifty percent wireless rates. If you are in
Italy and you need to use the web, voice browsing is becoming
the most common option. "The phone is the great social
equalizer," says Kris Hopkins, CEO of NewFound, a company
that offers voice browsing for search engines over wireless
devices. "Everybody has a phone and everybody will have
the ability to get on the Web." In Asia people are buying
phones over PC's at a 9 to 1 rate.
Level Three: The Voice Web
The next big step in voice technology is the voice web -
an entirely voice-based network of sites. To increase interest
in voice browsing and speech recognition several companies
have introduced forums for programmers to set up voice sites.
These forums become voice webs, sprouting voice auction sites
and voice based chat rooms. Nuance, BeVocal, and TellMe have
all introduced forums. TellMe, builder of web-based voice
applications for companies, pioneered the developers' forum
with the TellMe Studio. At the studio programmers have their
voice applications posted for free. Call up 1-800-555-TELL
to voice-browse through the creations of 10,000 developers.
BeVocal has its own active developer's program, the BeVocal
Cafe, where third parties develop applications that BeVocal
can host for their carrier customers (games, supposedly, are
in big demand). In September Nuance, the premier speech recognition
software manufacturer, waived its $495 membership fee for
its developers network and has new members joining at a rate
of 20 per day. Steve Elrich, spokesperson of Nuance, calls
this, "The birth of the internet all over again."
In the spirit of creating a voice community SpeechWorks has
launched its Open Speech Web, with an open source voice browser
based on the Linux model.
Voice Browsing Technology
The programming language responsible for connecting voicemail,
live agents, and speech enabled sites is called VoiceXML,
which was devised by Lucent, IBM, Motorola, and AT&T.
Although there is one defacto programming language the software
that converts text to speech and back varies. IBM has several
patents on its natural language understanding engine which
uses probability to guess what people mean if the words are
unclear. SpeechWorks technology is based on research conducted
at Massachusetts Institute of Technology. Phillips offers
Speech Pearl with a 200,000 word vocabulary, and Speechwave
offers text to speech in 35 languages. The leader in this
field is Nuance, working with around 1,500 application companies
and 1,500 platform companies.
For another programming option VocalPoint converts regular
HTML text into a voice platform. "If you've got a HTML
structure on your web site - you've got a platform for our
voiceplatform," says Kurt Losert, their CEO. VocalPoint
needs a day to set up and about 10 hours of programming. VoiceXML
takes 15 days to optimize a web site and over a hundred and
fifty hours of programming hours.
How well all this technology works depends who you are talking
to. Yahoo-by-Voice claims to be a voice portal but it requires
touch-tone responses. Much better are the technologies like
IBM's natural voice understanding engine, which converts complete
sentences and can even ask for missing information. SpeechWorks
and United Airlines recorded a 97%-99% accuracy reading for
their service. The most fun I had with voice browsing was
calling Phillips Speech Processing center and talking to their
browser which did not recognize the words, "Public Relations,"
"Media Relations," or "Press." Finally,
I asked for the operator.
C.J. Kennedy is working on voice-recognition software
for toasters.
|