Speech Recognition software (Dragon)

[pnorman]

That's supposed to have a very high accuracy rate, with 99.3% quoted in the write ups. That's of course with an analog signal from the mic., feeding into the sound card. But would it be likely to work as well with a signal that had once been digitally compressed? I'm thinking of all the books that I now have in Audible.com format. I occasionally "quote excerpts" from them, and it would be very handy to do that by just feeding my media device directly into the sound card.

Does anyone here know anything about such software?

pnorman

[Why Syzygy]

1. Do you think it's more accurate than the windows

voice recognition? You could try it with that before spending the money. All I know is years ago I did 'chat' with a guy using Dragon, and he had to re-type a LOT of his messages. It was incomprehensible. Surely it has improved since then. And iirc, the programs do have the ability to self correct after some experience. That is, the recognition increases with your input.

[pnorman]

2. I hve no idea of the accuracy of the Windows one (I didn't even know it existed!)

But I guess I could search it out and try it on digitally compressed voice files. I have no need for voice recognition software per se. But if it'll work okay on such voice files, $30 may not be such an extravagance.

pnormn

[Why Syzygy]

3. Price has come way down.

Look under C:\Program Files\Common Files\Microsoft Shared (XP)
It shows in my Add or Remove Programs. I used it with Word a bit.

[Fumesucker]

4. You might try this..

http://jott.com/default.aspx

With Jott Assistant and Jott for Salesforce, you just call a simple phone number, speak your notes, messages, or updates and hang up. Jott Voicemail works the same way, but your friends, family, and colleagues are the ones leaving the messages when they call your number. Then, Jott takes the spoken messages, turns them into text, and sends them to the right destination via email, text message, or web update.

[Jersey Devil]

5. If you're from Joisey, fogeddaboudit!

LOL, tried IBM Via Voice way back when and don't know if it was due to bad programming or my heavy joisey accent, but I had to correct and retype just about everything I dictated.

[RoyGBiv]

6. Voice recognition relies on learning ...

For the software to achieve anything near that rate of accuracy, it has to "learn" your voice. This is done simply through repetition.

The way Dragon's software works is to have to speak a few sentences/paragraphs that it displays for you, and it learns your initial patterns from that. Afterward, as you use the program, you make corrections and refinements that allow it to understand your specific speech patterns.

There is no way any recognition software is going to reach that level of accuracy on a recorded voice for which it has not been trained, unless the words being used are simple and predictable.

[pnorman]

7. That's the answer I was lookig for Thanks!

I knew about all the teaching sessions required, but had completely forgotten about it!

pnorman