I just started messing around with some speech transcription software this evening... this package is supposed to be pretty good, but I would assume that that's after you've gotten it properly and fully trained. I've done some preliminary training, but not the full set of training yet... so at this point, the software is still making some rather amusing errors. Just for kicks, I dictated a letter to my sweet babboo and sent it to her without making any corrections. At her request, I'm posting it here. The sign off was what really got us both roaring. I typically sign my letters to her "Your Imzadi now and always". I didn't expect MacSpeech to know "Imzadi", of course, but the mangling was still quite a hoot.
[begin]
Hello, sweetheart...
I'm for him around with MacSpeech this evening. Definitely slow going. But on the other hand, I have to admit that speech recognition these days is definitely better than was when I was in college! With very little training, it's already doing quite well as you can see, although there are minor errors here in there.
There are several training texts that you read to MacSpeech training to recognize your voice better. They're not a dozen or so. I've read only to them, so maybe they'll start working better after I have done the rest of them.
Just a short note. They go back and correct all this, but of course, part of the point is descended to you so you can laugh at it! To see whether you can hear what I was actually saying! Man, some of these mistakes are pretty bad...
I love you, sweetheart... (now, let's see how it tries to trends right mind that name for you!)
You or the Sunday now I'm always,
Parish
[end]
What's scary is that my brain keeps insisting that it's all proper English and that if I read it over-and-over again often enough, it'll magically start making sense. *eyes start to cross*
What's even scarier: the reason my brain is certain that it's "proper English" is probably because it's exactly what genuinely coherent language seems like to me. If I'm alert enough, my brain can "translate" almost in real-time for written language -- but no matter how awake I am, speech input usually requires conscious translation.
Posted by: Moggy at February 22, 2004 12:39 AM