I'm trying to convert to this stuff and it's driving me bats.
The chief problem is accuracy. The program, Dragon
Naturally Speaking, makes lots of mistakes so that
it takes a long long time to write anything, lots
of corrections. I've been at it for a couple of weeks.
Anybody have experience? Does this slowly get
better? Does this stuff ever work? Best, Jim
voice activated software
- Tom Dowling
- Posts: 511
- Joined: Thu Jun 28, 2001 6:00 pm
- antispam: No
- Please enter the next number in sequence: 8
- Tell us something.: Well, I've been a contributor and visitor to this site since 2001. At one time or another, one of my photographs was the opening page photograph. My teacher was Bill Ochs. I play the Penny Whistle. Not a lot else to say.
- Location: Brooklyn, N.Y.
Hello Jim:
IMHO, there are two technologies that have been slow to be perfected and made easy to use: voice recognition to text software and, to a lesser degree, Optical Character Recognition (w/ the major exception of the banking industry).
FYI, my podiatrist has a small PC and dictation set-up in each of his examining rooms, and he reviews your file on line and adds to it by dictating as the visit proceeds. The relevant notes from your visit are on the screen and in your file before you get out of your seat. He only recently 'upgraded to the paperless office' and, at first, it was driving him nuts. However, as time has passed--I've been getting treated for about 2.5 months--he is getting more productive with it. So, I guess there is a steep learning curve. I will certainly not miss the circumstance that has enabled me to do this "research".
Given your prodigious text output, as for example, on some of the more enduring threads on C & F, I wish you the best. Most importantly, will you be able to maintain those short pithy lines when your text input becomes all voice activated? Does the software recognize some sort of sotto voce commands for carriage returns, etc., so that it does NOT turn into text when you issue a verbal 'meta-command' (like Bold, Underline, etc)?
Good Luck!
Tom D.
IMHO, there are two technologies that have been slow to be perfected and made easy to use: voice recognition to text software and, to a lesser degree, Optical Character Recognition (w/ the major exception of the banking industry).
FYI, my podiatrist has a small PC and dictation set-up in each of his examining rooms, and he reviews your file on line and adds to it by dictating as the visit proceeds. The relevant notes from your visit are on the screen and in your file before you get out of your seat. He only recently 'upgraded to the paperless office' and, at first, it was driving him nuts. However, as time has passed--I've been getting treated for about 2.5 months--he is getting more productive with it. So, I guess there is a steep learning curve. I will certainly not miss the circumstance that has enabled me to do this "research".
Given your prodigious text output, as for example, on some of the more enduring threads on C & F, I wish you the best. Most importantly, will you be able to maintain those short pithy lines when your text input becomes all voice activated? Does the software recognize some sort of sotto voce commands for carriage returns, etc., so that it does NOT turn into text when you issue a verbal 'meta-command' (like Bold, Underline, etc)?
Good Luck!
Tom D.
- CTWhistler
- Posts: 22
- Joined: Sun Aug 17, 2003 9:04 am
Hi Jim,
Doesn't DD have a feature where it continues "training"? It's been a while since I used it, but I recall either adding new words to its databank when I tripped it up or running extra training sessions with the text that comes with the software. It did get better over time, but I still never took to it.
How good is your mike? Could that be part of the problem.
Tery
Doesn't DD have a feature where it continues "training"? It's been a while since I used it, but I recall either adding new words to its databank when I tripped it up or running extra training sessions with the text that comes with the software. It did get better over time, but I still never took to it.
How good is your mike? Could that be part of the problem.
Tery
-
- Posts: 243
- Joined: Thu Oct 25, 2001 6:00 pm
- antispam: No
- Please enter the next number in sequence: 8
- Tell us something.: I haven't been active on the site for years. I'm busy raising grandkids so I don't get out to play much.
- Location: Pleasanton, Ca
I've used Dragon for 4 years and I get better than 95% accuracy. I dictate all my reports for work and email them into the office from home. My company issued me the computer and the microphone, although I can only use it for work. (That's why I'm typing this.)
The microphone makes a big difference. You want to use a headset. I've been using an Andrea anti-noise anc-700.
Did you do the training? Also, use the correction box since that improves recognition. If you just make corrections on the document, Dragon treats the recognition as correct and will repeat it next time. If you use a lot of technical terms, feed documents into the program and it will add the words to the custom dictionary.
The more powerful the computer, the better the results up to a point. The machine I'm using has a 2 gig pentium. Check the options and see where the speed vs accuracy slider is set. Set the memory for the recognition files to at least 40 megs, preferably 80.
The longer you use it, the better the recognition gets. It keeps a statistical database of words that you use and selects from similar sounding words based on which ones you use the most.
Most important, avoid speaking single words. The recognition is much better for phrases or complete sentences. I turned off the automatic punctuation. It kept adding commas everytime I stopped to think.
Angelo
The microphone makes a big difference. You want to use a headset. I've been using an Andrea anti-noise anc-700.
Did you do the training? Also, use the correction box since that improves recognition. If you just make corrections on the document, Dragon treats the recognition as correct and will repeat it next time. If you use a lot of technical terms, feed documents into the program and it will add the words to the custom dictionary.
The more powerful the computer, the better the results up to a point. The machine I'm using has a 2 gig pentium. Check the options and see where the speed vs accuracy slider is set. Set the memory for the recognition files to at least 40 megs, preferably 80.
The longer you use it, the better the recognition gets. It keeps a statistical database of words that you use and selects from similar sounding words based on which ones you use the most.
Most important, avoid speaking single words. The recognition is much better for phrases or complete sentences. I turned off the automatic punctuation. It kept adding commas everytime I stopped to think.
Angelo
Eat well, drink well, laugh loud and often.
- fancypiper
- Posts: 2162
- Joined: Mon Jun 02, 2003 1:08 pm
- antispam: No
- Please enter the next number in sequence: 12
- Location: Sparta NC
- Contact:
Have you checked out IBM's ViaVoice?
All the voice rec software has to be taught how to hear the operator.
Good mikes are a good idea as well I could teach mine much easier and faster with my AKG 3000 C than the el-cheapo stick mike sold in computer stores.
One of these days, I am going to have to try it in Linux, but it is just so keyboard friendly, it's easier for me.
All the voice rec software has to be taught how to hear the operator.
Good mikes are a good idea as well I could teach mine much easier and faster with my AKG 3000 C than the el-cheapo stick mike sold in computer stores.
One of these days, I am going to have to try it in Linux, but it is just so keyboard friendly, it's easier for me.
- buddhu
- Posts: 4092
- Joined: Tue Sep 23, 2003 3:14 pm
- Please enter the next number in sequence: 1
- Location: In a ditch, just down the road from the pub
- Contact:
Speech recognition software performs surprisingly well *after* the user has been training it for a while. The more time you invest in teaching your program how your voice says stuff the better it gets.
The headset mike is a must. Background noise will screw the performance every time.
All that aside, I gave up in the end and went back to my patent four-finger typing technique. Simpler and about as fast after proofing and correcting.
The headset mike is a must. Background noise will screw the performance every time.
All that aside, I gave up in the end and went back to my patent four-finger typing technique. Simpler and about as fast after proofing and correcting.
And whether the blood be highland, lowland or no.
And whether the skin be black or white as the snow.
Of kith and of kin we are one, be it right, be it wrong.
As long as our hearts beat true to the lilt of a song.
And whether the skin be black or white as the snow.
Of kith and of kin we are one, be it right, be it wrong.
As long as our hearts beat true to the lilt of a song.
- burnsbyrne
- Posts: 1345
- Joined: Thu Apr 11, 2002 6:00 pm
- Please enter the next number in sequence: 1
- Location: Cleveland, Ohio
About 4 years ago I tried using IBM ViaVoice because my arm pain was acting up badly. I had done a trial with Dragon but I was put off by the price. I found ViaVoice worked well for me for screen and cursor commands like "next"(screen), previous, up, down, end, etc. I tried using it for dictation but I found I had to make so many corrections that it was less work just to type the text myself than to correct all the errors. However, I didn't give VV a lengthy trial because I moved on to another project which required less typing. The program also may have been improved in the meantime so YMMV.
Mike
Mike
- antstastegood
- Posts: 519
- Joined: Wed Apr 23, 2003 12:48 am
- Please enter the next number in sequence: 1
- Location: Seabiscuit's stomping ground.
- Contact: