Brave New Yackety-Yak.
I am not a computer jock. I don't even remember how much memory I have on my PC. But when I began reading about voice recognition software, the technology that allows computers to translate audio input into written text, I was intrigued. If it really worked, this very 21st-century technology could open doors for people whose disabilities stop them from using a regular computer keyboard. I decided to try it out myself.
The availability of voice recognition computing can be attributed to advances in telecommunications technology meeting advances in home use hardware. These programs require a significant amount of memory (RAM) and speed (MHz). Fortunately, most home PCs on the market are adequate.
Voice recognition has now moved beyond the original "discreet speech" mode, when YOU. HAD. TO. SPEAK. SLOWLY. WORD. BY. WORD. The newer "continuous speech" programs require you to speak naturally, in complete phrases or sentences. The programs listen to your vocabulary, and also to your syntax, cadence, and particular grammatical structures. You don't have to pause between every word anymore, but you do have to enunciate--and that takes a certain amount of concentration.
IBM's ViaVoice ads have made continuous speech software familiar to many TV viewers. IBM's main competitors are Dragon Systems Inc. and Lernout & Hauspie. In general, the IBM products are cheaper and a little easier to use with other applications. Lernout & Hauspie's Kurzweil Voice Xpress Plus program is easy to learn, but its recognition time is slower than the other programs. Dragon's programs tend to be recommend over the others since they are the most accurate and user-friendly, though they eat up more computer memory.
Do a little consumer research, and you'll find a daunting array of products to choose from, with prices ranging from under $100 to nearly $700. Bigger is not always necessary. Before deciding which software program to buy, figure out what you want to do with it.
I've been trying out Dragon System's NaturallySpeaking Personal Edition 1.0 program ($99) for the past 3 months. I can now open the Dictate program, blab away, format the text using dictated commands, and cut and paste into another window such as Word or e-mail--without touching my keyboard. Since I don't usually work with many different applications, I am perfectly happy with this program.
Joe Karnicky, a full, time computer programmer who works from his home in Menlo Park, California, began using an early version of the Dragon Dictate program 6 years ago. His MS has limited his keyboard use to one finger, but he can use the mouse. Since he frequently switches among 5 and 6 applications as he works, the command-and-control feature suits him just fine. (This command-and-control feature is now available, along with the continuous speech feature, in the Deluxe Edition program for $695.) "I don't know what I would do without the program," said Mr. Karnicky. He hasn't bothered looking into the newer editions of the software.
Linda Storey of Englewood, Colorado, was a rock and roll musician who knew music, not computers. Progressive MS eventually made it impossible for her to play guitar in a band. She got her PC and an older version of the Dragon software through a local grants program. This technology has made it possible for her to write, produce, and distribute a CD of her own songs.
The songs on Linda Storey's CD, "Willow Tree", deal with her feelings about assisted suicide, living with MS, and her strong Christian faith. Since she has little arm strength, she composes on a keyboard using a pencil in her mouth to hit the notes. Then she dictates her song lyrics into the computer.
She is working on distribution using her voice recognition software to write necessary letters or send faxes. Some Christian Rock radio stations are now playing her CD, so Mrs. Storey got hooked up to the Internet and had someone design a web page for her business.(*)
Lawyer Jeff Crosby started out with the NaturallySpeaking Deluxe program this past year at his Lancaster, Pennsylvania, firm. Mr. Crosby joined the firm in 1981, shortly after being diagnosed with progressive MS. The scope of his job has changed from doing things like special education litigation to doing transactional jobs, such as real estate contracts. He now uses a motorized wheelchair, and finds it difficult to use the computer keyboard. Much of Mr. Crosby's initial frustration with his new software disappeared when he upgraded his computer. And while he still gives dictation tapes to his secretary to transcribe, the NaturallySpeaking Deluxe program allows him to read and send interoffice e-mail, generate his own documents, and access shared office files. The Deluxe program has also allowed him to create macros, those shortcuts that condense a 5 or 6 step process into one keystroke ... or one voice command, in this case.
Getting started's hard to do
Once I got the audio configured so that my microphone worked properly, I completed the hour-long initial NaturallySpeaking voice training. I read out loud, while the program matched the way I spoke against the selected text, creating my voice template or profile. Every time I use the program now, I have the option of updating and fine-tuning this profile--which is what builds accuracy rate.
I'm so used to typing on a keyboard and navigating with my mouse that I found myself tongue-tied when I finally arrived at the blinking "microphone on" icon. It's strange to speak your thoughts, and then see them appear as screen text. My initial embarrassment disappeared fairly quickly, though, when I saw that the words on the screen often bore no resemblance to what I thought I had just said.
"A now trying at the new voice recognition program jungle in the writing article about."
Huh? After generating a few sentences, I would read back, only to find that I couldn't remember what I had just tried to say!
Scratch flint ...
Luckily, the program puts many voice commands at your disposal. For instance, I could say "Select `A now'". These words would be highlighted, and then I could speak the correction, "I am now". There are commands for formatting, moving through the document, and editing. You can also have your computer play back all or parts of the text. The phrase "scratch that" deletes an erroneous word or phrase right after saying it; anyone overhearing my first few forays into NaturallySpeaking might have thought I had a bad case of poison ivy.
After these first sessions, I felt incredibly mealy-mouthed and inarticulate. It was frustrating to have to pause after each sentence or phrase to make corrections. Sometimes I'd have to open the "correct spelling" window when the machine simply wouldn't--or couldn't--"hear" what I was trying to say. I also found that I frequently needed to take the time to open the "train" window (its icon is a seal balancing a ball on its nose). Here, you say one particular word, and the program adds this piece of information to your speech file.
Dr. Mark Griffith, of Zephyr TEC, works with vocational rehabilitation programs in Oregon and Washington to train people in the use of voice recognition software. "Users expect too much, too fast," he cautions. Speech recognition software may be flying off the shelves, but it's also the software product that gets returned the most frequently He advises people that they will need a minimum of 20-40 hours of basic use and training. It took me a whole week of painstaking work before I was able to use the program with some degree of accuracy.
Many ways to skin a dragon ...
The Dragon NaturallySpeaking software is basically set up to run with Windows. However, Valerie Wilson of Lancaster, Pennsylvania has been successfully operating her software through DOS. Ms. Wilson has progressive MS, and is quadriplegic. With 2 daughters headed for college, she urgently wanted to get back to work. She entered a vocational training program provided by the United Cerebral Palsy Foundation. Even though Ms. Wilson was an experienced secretary, the task of mastering a new technology and the skills to go along with it took time and commitment. Ms. Wilson now says, looking back on her training, "You just have to use the technology--you can't be afraid of it."
Because of it, Ms. Wilson is working at home transcribing medical tapes. She wears earphones, and presses the Play and Rewind buttons on the tape recorder with her forehead. The DragonDictate program allows her to "talk" the tape contents into text. Her old PC is slow, but she's working 24 hours a week. This year, for the first time in a long time, she had the thrill of paying taxes on her earned income!
Ms. Wilson sent out over 90 application letters before landing her present job. "We can do the work," she says, thinking about some employers' reluctance to hire people with disabilities. "We just go about it a little differently."
More than one voice
I put on my head-mike and loaded up the NaturallySpeaking program one morning before my first cup of coffee to find out what the computer would make of my croaky morning voice.
"I just wanted to see what would happen speaking this early on morning when McCoy says of oven of a fraud at."
Not too bad. (What I said was "I just wanted to see what would happen speaking this early in the morning when my voice has a frog in it.") Apres coffee was much more successful.
Some people with MS experience slurred speech as a symptom, and this could present a problem to voice recognition programs. If speech is slurred most of the time, the user would simply train the computer to recognize this voice. But if speech becomes slurred intermittently or when the person is fatigued--a typical MS story--the program won't recognize what is said.
There is a solution to this, though not an easy one. Several software programs allow for multiple voice profiles. Someone with intermittently slurred speech, or with allergies for that matter, can maintain 2 voice profiles, and switch between them as necessary. The downside is that all corrections or fine-tuning made for one profile have to be replicated in the other.
Talking the talk
The new voice recognition programs are being used by a wide variety of people: data entry people with carpal tunnel syndrome; fiction writers experimenting with dictation as a creative mode; CEOs who never learned to type; and people with physical disabilities who would not otherwise have access to a computer or a job.
The people I spoke with for this article were using different voice recognition programs to do many different things, professionally and on a personal level. If you think that this software might be useful for you, first speak with a professional consultant or vendor to clarify your goals. The major software manufacturers list some vendors on their web sites. You can also check whether your nearest Vocational Rehab Center offers software training of this sort. Second--have patience. I know how long it took me!
(*) "Willow Tree" is available on CD for $15, and on cassette for $12, and may be ordered from: Linda Storey Publishing, P.O. Box 1784, Englewood, CO 80150. Or order online: www.LindaStore.com.
RELATED ARTICLE: Speech Recognition Tips
The following list can help you set up your speech software for maximum performance and save you a fistful of hair if you can't seem to get your system to work the way it does on the TV commercials where "you talk and it types".
* Typically you need to have at least a Pentium 166 (MHz) with 32MB of RAM and about 65MB of hard drive space. You will need more RAM if you are running WinNT. Buy the best PC you can afford.
* All speech recognition products list compatible soundcards either on the software box, in their manuals, or on their web sites. Play it safe and make sure you have an industry-standard 16-or 32-bit soundcard, or a built-in sound system that is "Sound Blaster" compatible.
* Many of the stand-up microphones that are shipped with new PCs will not work with speech recognition software. Most speech software is shipped with a microphone that will deliver excellent quality sound.
* To test your microphone, first record your voice to the Windows Sound Recorder. If the playback is loud and clear, you probably won't have any problems with recognition. If it is fuzzy, muted, or otherwise distorted, check the rest of this list before you install your software. [Note: the Windows Sound Recorder can be found in Start/Programs/Accessories/Multimedia.]
* If you can't get a clean recording in the Windows Sound Recorder, affix the microphone cord to the bottom of your desk with tape so that it is not touching any other components or cords. This can eliminate electrical interference that can distort your audio input.
Improving Recognition after Installation
* Moving your microphone just one inch can throw your recognition rate out the window. For best results, place the microphone about one thumb length away from and to the side of your mouth.
* The speech engine will perform best if you are speaking with careful enunciation. If other people can understand you, chances are so can the speech engine.
* The better speech recognition products use syntax and pace as part of the algorithms for recognizing your speech, so you will get higher accuracy if you dictate in complete thoughts and sentences.
Help is Available
Whether you are a first-time computer user or a technology specialist, voice recognition can be tricky to set up and even trickier to learn. Don't fret. Most speech recognition products come with optional multimedia tutorials and well-documented manuals. Both IBM and Dragon also offer online help that is easy to navigate and always available.
If you will be using your speech recognition software at your workplace, you may want to utilize a consultant familiar with both voice recognition software and your application suite. A good consultant can help you build productivity-enhancing tools so that your software delivers business results for a long time to come. Both IBM and Dragon offer referrals to consultants nationwide.
A. Erik Chesla is a speech recognition consultant for Quantum Leap Systems in New York City. He specializes in technology solutions for people with physical disabilities. He can be reached at email@example.com or 212-473-9060.
Lorna Smedman is managing editor of this publication.
|Printer friendly Cite/link Email Feedback|
|Date:||Jan 1, 1999|
|Previous Article:||About Redefining My Fight Against MS.|
|Next Article:||WHEN THEY'RE STANDING AND YOU'RE NOT.|
|Yackety-yak about race.|
|A human migration fueled by dung?|