Speech systems reduce the need for keyboards

Speech systems reduce the need for keyboards<@VM>You can train speech recognition software

BY JOHN MCCORMICK | SPECIAL TO GCN

I have been testing and using sound recognition and speech recognition hardware and software since 1982.

The early systems worked by speech training, meaning that they did not work at all until I trained them. But because they could only recognize a few hundred words, they could be surprisingly accurate and useful in controlling my computing environment.

I have recently used two low-end speaker-independent speech recognition programs, IBM Corp.'s ViaVoice Standard and a version of Lernout & Hauspie's Dragon NaturallySpeaking that is similar to NaturallySpeaking Essentials, but it came bundled with other systems and is slightly different from the retail product.

Keyboard reliance

Of the two, ViaVoice is less useful for severely disabled users because it relies on keyboard or mouse cursor control and screen navigation. Using those controls is an advantage for people who can use navigation hardware'it reduces the number of commands ViaVoice has to learn and typically is faster than using voice commands. But it limits what many disabled users can do.

More advanced versions of ViaVoice now include many voice navigation capabilities.

ViaVoice also includes text-to-speech conversion, which can be fine-tuned to create a relatively pleasant, or at least tolerable, electronic voice.

Dragon NaturallySpeaking can operate as a hands-off system and, in addition to taking dictation, can navigate documents, change fonts, edit, modify paragraph alignment and even create bulleted lists.

Both low-end programs do well on discrete speech, when each word is spoken between distinct pauses, and both do fairly well right out of the box with continuous speech.

More important, any speech recognition program will become easier to use and more accurate with time because the software can be trained.

Both of these programs are speaker-independent. That is, they will work as initially installed for any English-language user who speaks clearly and consistently.

Using either one of these programs puts a strain on a user, who must always speak in a controlled, even tone or jeopardize accuracy. I have tried, without notable success, to train older individuals to use these programs rather than have them try to learn touch typing. Each one found it difficult to rely on speech recognition. They usually were able to outperform it using hunt-and-peck, one-finger typing'especially because corrections to the speech recognition had to be typed in anyway and error rates were high.

I have concluded that to use speech recognition successfully in its present form, a user must be already entirely comfortable with computers.

That doesn't mean it isn't practical. A motivated person with severe motor control impairment could, with practice, easily achieve 20 words per minute or faster of text input. People who are really comfortable with speech recognition do even better, although slower computers will limit recognition speeds.

Accuracy is a concern

I tested the low-end version of each product. There are more sophisticated editions with many valuable features, but they aren't noticeably more accurate when it comes to word recognition.

Here is a sample of my first attempt to use ViaVoice after a few minutes of training:
'Those the blind folded them to come to the the the party'

Do you recognize the typist's standard test sentence: 'Now is the time for all good men to come to the aid of their party'?

But using the training/correction mode, a few minutes later this was the result:

'Know is the time for all good men to come to the aid of their party.'

That's pretty close after only a few minutes of use, and it's a good indication that even the least expensive version of ViaVoice has real promise for motor-impaired users. And it is included with some IBM computers.









































































































Vendor Product Platform Notes Price
IBM Corp.

Armonk, N.Y.

914-499-1900

www.software.ibm.com
ViaVoice Standard Win95 up Basic speech recognition $50
ViaVoice Advanced Win98 up Includes large vocabulary, Web navigation, USB microphone $90
ViaVoice Pro Win95 up Includes large vocabulary and powerful navigation tools $189
ViaVoice Pro Elite Win95 up Uses desktop microphone instead of headset $249
ViaVoice Mac Enhanced Mac OS Includes USB microphone $141
Lernout & Hauspie

Burlington, Mass.

781-203-5000

www.lhsl.com
Voice Express Standard Win95 up Basic speech recognition $50
Voice Express Advanced Win95 up Includes Word and WordPerfect command support $80
Voice Express Mobile
Professional
Win95 up Includes Olympus digital voice recorder for dictation $230
Voice Express Professional Win95 up Includes command control for Microsoft Office, including PowerPoint and Excel, along with speech synthesis $149
Dragon NaturallySpeaking
Essentials
Windows Basic speech recognition includes e-mail and Web surfing support $55
Dragon NaturallySpeaking
Standard
Windows Adds more sophisticated editing features $100
Dragon NaturallySpeaking
Preferred
Windows Includes full-featured editing and navigation support for Word and WordPerfect $169
Dragon NaturallySpeaking
Preferred USB
Windows Includes USB microphone $210
Dragon NaturallySpeaking
Professional Medical
Windows Includes comprehensive
medical vocabulary
$730
Dragon NaturallySpeaking
Professional Legal
Windows Includes comprehensive
legal vocabulary
$690
Dragon NaturallySpeaking
Professional Public Safety
Windows Includes comprehensive
public safety vocabulary, including profanity, acronyms
$570


inside gcn

  • A framework for secure software

Reader Comments

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above

More from 1105 Public Sector Media Group