Home

Taming a Dragon

« previous entry | next entry »
Apr. 30th, 2008 | 12:55 pm

I blogged, in passing, about the difficulties I have using the voice recognition software, Dragon 9.1.

lenrosen has expressed some surprise that I have had so much trouble with the program, citing only a 2% error rate.

I have never come close to that. I’ve never tried to calculate it, but my error rate must be well over 20%, leading to a lot of frustration.

It’s not a matter for harbinger [[[ example: what I said was ‘it’s not a matter of poor hardware’ ]]]

Starting again: it’s not a matter of poor hardware. I have a good computer, with a fast CPU, and lots of memory. I use a good quality headset, not the one that came with Dragon.

My post-stroke speech is not perfect, but it’s clear enough.

I would love to have a 2% error rate. Maybe by the time Dragon 10 or 11 comes out...

Then, perhaps, it will be like Star Trek, where you can say:
“Computer, connect the atmasfras to the Heisenberg compensator...”
with nary an error.

Link | Leave a comment | Add to Memories | Tell a Friend

Comments {2}

Mike Toot

(no subject)

from: [info]miketo
date: May. 1st, 2008 03:19 pm (UTC)
Link

I love the *concept* of voice recognition software and have twice experimented with earlier versions of Dragon (v.3 and v.5). Each time, I went back to typing. Even though I'm semi-classically trained in speaking (theater major, lots of Shakespeare, the usual drama nerd stuff) and do training and customer presentations, I still had significant error rates. Using the voice commands to correct the mistakes took far too long and introduced their own set of errors.

If you have the ability to evaluate ViaVoice, you could give that a shot and see if its algorithms do a better job for you.

Reply | Thread

rdeck

(no subject)

from: [info]rdeck
date: May. 1st, 2008 09:50 pm (UTC)
Link

I'm not familiar with ViaVoice. I'll look into it.
Thanks

Reply | Parent | Thread