Voice Recognition Programs Anyone ??

[Richard_Cranium72]

http://www.us.buy.com/retail/software/subcat.asp?loc=2080

A retired pal of mine is considering getting a pewter.. However he don't type and wants to know the ups and downs of this type software..

i.e. Does it work ?

thanx

DrVette

[big_block_buick]

ive known 2 people who have tried dragon naturally speaking ,and while i cant say thumbs up or down they dont currently have it loaded on their systems..but they just used it as a toy...you have to speak clearly for like hours on hours to train it to recognize your voice..and when i seen it in in action(after it was supposedly trained it was messing up a few words here and there..sorry not much help...

[Brangwen]

I've been using Dragon Naturally Speaking for approximately 4 years. It does not require very much time to "train" it to recognize one's voice. When reinstalling it on a new HD, for example, I'll dictate for an hour and it will transcribe my words to the monitor fairly accurately. It takes a while to learn how to use it. For example, you cannot slur your words, i.e., you must enunciate to a degree. A user has to take breaks after dictating for a given amount of time, depending on the person, as a user will begin to slur words from fatigue which the program picks up though the user does not. Most dictation will require proofing. Sometimes, for example, I'll dictate "I am "being" a good guy." The word "being" may be recorded as "bean." In addition, speech models will be built over time. This means that the user, by saving speech files after using the software, and by loading several sample written documents that reflect the user's writing style into the program, the software will "learn" how the user customarily constructs phrases. So when in "doubt," the software will reference these models and "assume" which word a slurring user intends to include in a particular word pattern and use it. Also one can train the software to create macros. For example, I often say "tois" (pronounced "twah" and French for 3) and my software triggers a macro that creates three spaces. I happen to like three spaces before the beginning of each sentence. Another example is creating a macro to insert a letterhead at the beginning of a letter and triggering the macro by saying "let-head." One can create as many macros as desired, and the word database can be expanded incredibly. I believe my program came with a 35,000 word database.

A very important point is the microphone must be positioned properly ALWAYS relative to the mouth. If too close and you exhale hard or breathe directly into it, the program will pick that noise up and attempt to translate it into a word. Additionally, it's not a good idea to stare at the monitor as you dictate, as it makes for a real unnatural way of speaking. I usually put together an outline of topics that I want to dictate and do so in as flowing a manner as possible WITHOUT looking at the monitor. I usually look at birds on a feeder outside my window or some other scenery. If you speak slow and choppy, the software has problems.

I primarily dictate directly into Corel Word Perfect. I believe there are features with Dragon that make it even more compatible with MS Word, which I do not care for.

I recommend this product if you want to dictate fairly long letters. Many times I'm just too tired to type and I enjoy kicking back and simply dictating.

I originally paid around $800 for the Deluxe version. That was four or five years ago. Since then I've switched to Professional which I believe is several hundred dollars less. I've read reviews on different speech recognition programs and IBM's ViaVoice (I believe that's the title) received favorable ratings and it's around $100. Also with Dragon, one can dictate into a recorder (say in the car) then plug the recorder into your "big" computer (desktop) and the program will transcribe the recording. Again, it will require proofing, but it's a great timesaver if you're serious.

I hope I've shed a bit of light on this interface. Good luck.

Brangwen http://sysopt.earthweb.com/forum/wink.gif

[Ed_S]

Brangwen - I'm interested in this too, and you sound like the most experienced person I've run into.

Just to show us how much editing is required by an experienced user, would it be possible for you to dictate something into Word Perfect, then copy it here unedited???

Thanks!
Ed

[thekingofpain]

You need to check out our own Techjumpers ubb, it pertains to Voice dictation only, im sure Brangwen would be a MUCH appreciated member there, http://www.kisssys.com/cgi-bin/Ultimate.cgi?action=intro have a look...

[Brangwen]

To Ed and others interested in voice recognition software:

After reading your response to my reply in the form, I decided to honor your request by dictating a relatively short latter (letter) without any proofreading with my dragon voice recognition program called NaturallySpeaking. After having typed so much text last evening in my response by hand, I am not certain entirely what I’m going to “talk” about wild (while) dictating this. As I look about my dad and (den), I see many objects on the wall. These include framed diplomas, portraits, clocks of various sizes and origins and awards which I have received from various institutions and organizations. I must admit that I have stolen a glance at my monitor every few sentences in order to see whether or not by dictation is appearing acridly. (accurately) You’ll notice that several of the words are not accurate, hence the need for proofreading. Again, a (I) truly do not know what more to say at this time. Oh, one more thing, I did look at the new bulletin board about voice recognition software and I hope to have an opportunity to Pettit (put) some relevant information on it that may be helpful to others. I also hope to learn from others about voice recognition software. I hope I have been up (of) some assistance by providing this “sample.”

Explanation about inaccuracies in the above paragraph. As I mentioned in my “hand written” responds (response) of last evening, one tends to make errors wild (while) dictating when a bit fatigued. I have been working for approximately six hours and as a consequence I am a little bit tired and I am more than certain that I have been slower in (slurring) my words wild (while) dictating this. Hence, the inaccuracies. Inaccuracy notwithstanding, one can see the real advantage of using this type of software if one is able to dictate had (at) a rather last (fast) rated (rate of) speed as compared to attempting to type at the same rated (rate of) speed. Believe me, I am no speed typist. It is much easier for me to speed (speak) quickly than for me to type quickly. I want our (to) review what I’ve dictated and I shall put in parentheses the correct word or rather the word I intended to have the software “type” directly following each inaccuracy in order that you may get a better idea of how the software works. At this point, I am reverting to conventional typing, starting now!

Inaccuracies here are absolutely due to slurring my words. Many of the mistakes below you can likely tell are due to the similarity of the word intended and the resulting mistake, e.g., I said “accurately” and “acridly” was “typed.” Understand as well, and a point I did not mention last post, the first words I dictated after booting the program (Dragon) were “To Ed . . .” I usually begin dictating sessions by saying a line like “Now is the time for all good men to come to the aid of their parties.” This, in my experience, gets my lips into dictating mode as silly as that may sound. Also, I hemmed and hawed about what I was going to say. Best results are achieved when the user is able and does speak quickly by thinking ahead as to what he “will” say like chess moves in a chess game. Less chance of stumbling over words, believe it or not.

latter = letter
wild = while
dad and = den
acridly = accurately
a = I
up = of
responds = response
wild = while
slower in = slurring
wild = while
had = at
last = fast
rated = rate of
speed = speak
our = to

Brangwen http://sysopt.earthweb.com/forum/wink.gif


[This message has been edited by Brangwen (edited 11-18-2000).]

[GenePG]

The newest version of Dragon claims "minutes" of enrollment, 250,000 word vocabulary and the ability to dictate almost anywhere in Windows. However, I have been using IBM ViaVoice and DragonDictate for all four years now. Right now, the biggest obstacle to accuracy is microphone placement. One of the name microphone manufacturers, Andrea, sells an "array" microphone that is supposed to pick up your voice from about two feet away. The microphone array sits on top your monitor or some other convenient place. Also, in order to take advantage of the full capabilities of DragonDictateyou need a faster processor and more memory than recommended. I'm running 256KB RAM and an 800 MHz Athalon processor. I am soon going to upgrade to the next version of DragonDictate (version 5.0) and I expect my current hardware configuration to be just adequate.

Text input directly into Microsoft Internet Explorer sysop.com post reply screen.


[This message has been edited by GenePG (edited 11-18-2000).]

[Brangwen]

Gene, you make some excellent points in your reply above. I agree that microphone placement is incredibly important. Just for comparison, I am presently dictating directly into this web site. My previous reply was dictated into Corel WordPerfect and copied and pasted onto this web site. I am dictating with a computer that is utilizing a 600 MHz Intel Pentium 3 with 128 MB of PC 100 SDRAM. This is a vast improvement over my 333 MHz Celeron configuration! A friend of mine who also utilizes Dragon's NaturallySpeaking uses a computer more similar in configuration to yours and has tremendous success with this voice recognition program. Frankly, this is one of the first times (actually, this is the first time) that I have actually attempted to dictate directly into the sysop.com forum, and find that my program responds rather nicely. I would be pleased to answer any further questions about this particular voice recognition product.

Brangwen http://sysopt.earthweb.com/forum/wink.gif

PS: For those who read my "dictated" reply above, you may notice that there are far fewer mistakes in this response. Another peculiarity of this program is that its accuracy increases after it has "warmed up." In other words, since I dictated my last reply, I have had the dictation program minimized but not turned off. So I am not speaking into it "cold," as I did when I began dictating directly into WordPerfect.

[Ed_S]

Well, I'm impressed! The accuracy is MUCH better than I expected. And from what you said it usually does EVEN BETTER??

Previously, the only experience I'd had was when a friend showed me it at his house over a year ago. He'd only had it about two weeks at that point, in fact he'd "borrowed" the program to try to learn it. The output was horrible, so error strewn that it would have been quicker to type it from scratch. He gave up using it.

Certainly the software has improved since then, but it's obvious that the experience of the user is the key element.

Thanks much for the effort, I'll look into it further.

Ed

[daveleau]

I could get it to distinguish between pin/pen/in on my system. Took a little training. The newer version V4 of Dragon and V4 of Learnout & Hauspie take 15min to set up. V3 takes about 2-3 hours of reading to set up. I found them to be very affective and useful. Be aware of system req's. A 266 is the slowest and the more pawerful, the more accurate it will be.
Dave

[Graham]

Tried a demo of game commander (its not just for games), worked very well with no training, however dont think you can use it for text entry (could be wrong though), worth checking out.

G

[Richard_Cranium72]

Brangwen, U iz da man http://sysopt.earthweb.com/forum/smile.gif

Thanks to all who responded.

I'm considering setting this guy up with my IBM w/475 AMD 64RAM, 100FSB, 7200HD.

I hope this is enough processor to run the program, time will tell.

He won't have to pay if the machine is too wimpy to operate the program, perhaps a dose of 128 PC 100 RAM More will assist..

thanx
DrVette

[TechJumper]

When trying to discerne between conflicting tones like Pin and Pen, the core engine analyses the preceding sentence, using built in grammer specs, like "context clues" and makes its decision in that way. While this system is fool proof, for it to work all the time requires an equally as perfect voice recognition engine, that can inturpret the minute variances in a persons voice.
This kind of break-down, for the time being, also requires a very powerful microphone, and high-quality sound card.


the most accurate systems involving this area have been Dragon, especially release 5.0. I hope to get a trial sent to me to review on my site, where I will discuss issues like these, and how these nuances could affect consumer acceptance of this new technology.


Best,

Mike

[TechJumper]

I was reading that one of you has tried VR in a game?

One game that really incorporates this technology into their system is Microsoft GameVoice www.gamevoice.com. (http://www.gamevoice.com.)

This product clearly shows the future and potential of voice recognition.


Mike

[bdunn]

I use IBM Via Voice98 and it works fine for me except at times like now when I have a cold. I trained an additional user for occasions like this so I now have a seperate setting for when I am hoarse.