[If they can ever get this to work well in consumer settings it will be 
an important adaptive technology for the blind and the mobility impaired.]

Speech-recognition technology is rapidly improving

By Troy Wolverton
Mercury News Columnist

Posted: 07/17/2009 12:00:00 PM PDT

http://www.siliconvalley.com/opinion/ci_12854535


Maybe I watched too much "Star Trek" when I was younger, but I love the 
idea of being able to command things in my house or in my car by talking 
to them.

I'd love to be like the Enterprise's Captain Picard and get a cup of tea 
by walking up to a wall panel and saying "Earl Grey, hot."

I'm not holding out for a voice-activated replicator. I'd settle for 
simpler stuff, such as being able to set the thermostat, turn on my 
oven, switch off the lights or record a ballgame by simply speaking my 
wishes out loud.

Unfortunately, my real-life experience with speech recognition 
technology — and I'm guessing yours, too — has been nothing like "Star 
Trek." Instead, it's been largely frustrating and sometimes infuriating, 
what with voice-dialing cell phones that can't understand who you're 
asking them to call and phone trees that respond to simple voice 
commands but won't direct you to a live person who can handle more 
complex questions.

But if speech recognition advocates are to be believed, the science 
fiction world of widespread and well-functioning voice-controlled 
devices and appliances is finally just around the corner.

"I think we're seeing a real renaissance here," said Bill Meisel, a 
longtime follower of speech-recognition technology and editor of Speech 
Strategy News, an industry newsletter.

Meisel foresees a not-so-far-off world in which our mobile phones serve 
as a universal remote that allows us to control and program everything 
from TiVos to alarm clocks with our voice.

Todd Mozer, CEO of Sensory, a Sunnyvale-based speech-recognition 
company, envisions a soon-to-come world filled with speech-controlled 
Internet devices, or SCIDs.

One example of such a device might be a clock radio that could not only 
tell you the current time when asked, but could also, using its Internet 
connection, tell you the weather in Boston.

Regardless of how the future unfolds, advocates such as Meisel and Mozer 
say it's nearer than most people might think. Voice recognition is not 
some pipe dream, but a maturing technology that has improved in recent 
years and already works well in certain circumstances.

Widespread adoption of speech recognition has been delayed by problems 
with accuracy, advocates acknowledge. But thanks to Moore's law, 
computers and microprocessors are getting faster at processing spoken 
words. They're also able to sort through and compare what's spoken with 
increasingly large databases of recognized words and ways of saying them.

Developers are also improving accuracy by narrowing the problem.

Some do that by limiting the number of words a device needs to 
understand. A Bluetooth headset, say, may recognize only a handful of 
commands. But it will understand those commands so well that it can 
respond to them, no matter a person's accent and despite having to 
listen for them over a talk-radio program.

Another way of narrowing the problem is to categorize the information 
sought, or focus on industry-specific vocabularies. Some speech-to-text 
programs, for example, are specifically designed for particular medical 
professions, such as radiology.

But it's more than just improved accuracy that has advocates excited 
about prospects for speech-recognition. The technology has become a big 
business, and major corporations, such as Apple, Microsoft and Google, 
are pushing it. Speech recognition technologies have become commonplace 
not only in cell phones and phone trees, but also in particular 
industries, such as transcribing medical records or even for use by 
workers in larger warehouses.

And two other trends could help move speech technologies into consumers' 
homes. One is the growing demand for more "natural" ways of interacting 
with technology products, such as by using gestures on a touch-screen or 
via motion-sensing controllers. Some analysts think speech works well in 
tandem with other natural interfaces.

The other trend is the growing number of home networks and 
Internet-connected devices. Having a network connection allows devices 
to download updates that would improve their speech recognition 
capabilities over time. It also can let devices tap into more powerful 
speech recognizers either on the local network or out on the Internet.

Indeed, getting speech-controlled devices into the mainstream is 
becoming less an issue of technology than consumer acceptance, advocates 
say. The more consumers are exposed to speech recognition services and 
have positive experiences, the more likely they'll seek speech-based 
interfaces in other areas, argues Meisel.

"People's attitudes have been colored by unpleasant experiences," said 
Meisel. "(Those) attitudes will change."

Here's hoping he's right. Dreams of "Star Trek" aside, I'd be happy to 
simply have my phone recognize when I want to dial my sister.

-- 
================================
George Antunes, Political Science Dept
University of Houston; Houston, TX 77204 
Voice: 713-743-3923  Fax: 713-743-3927
Mail: antunes at uh dot edu

***********************************
* POST TO [email protected] *
***********************************

Medianews mailing list
[email protected]
http://lists.etskywarn.net/mailman/listinfo/medianews

Reply via email to