Long story short, I messed up my left shoulder. I’ll be seeing an orthopedist tomorrow, but it’s difficult to type for more than five minutes at a stretch. I can click a mouse and do a large portion of my job, but I do have an article and a couple of presentations to write in the next few weeks.
Anybody have experience using the MS speech to text program? It seems promising, but then what doesn’t on paper?
I don’t know anyone who has had any luck with it whatsoever. I get the distinct impression that Microsoft didn’t put much time or effort into developing it, but just wanted to claim that they’d developed such a program for disabled users.
Dragon Naturally Speaking, on the other hand, works great in versions 9+. Version 11 is in stores now, and you could probably get an older (and cheaper) version online within a few days.
The speech recognition (SR) engine in Windows Vista and 7 is pretty much state-of-the-art. I worked with SR and text-to-speech (TTS) a lot in my last job (and I am a co-inventor on a patent for improving SR). Compared to the various Nuance products [they own or have exclusive license for virtually every SR engine on the planet, including Dragon Naturally Speaking], the most recent generation MS engine have a higher success rate in dictionary mode. Microsoft Research does a lot with SR, and some of their work made it into Vista. What Microsoft does not have (as a commercial product) is a good server-based SR engine that can do recognition on voice segments, such as is required for IVR (telephony) systems, nor do they have a grammar-based engine (which generally produces much better results, but requires that the user keep to a very restricted vocabulary).
For dictation, such as document creation, the built-in Windows engine will give as good or better results than a third-party product. Ideally, the Windows engine would have a grammar-based engine for menu navigation and file and application control, but they don’t. If third-party products support grammar-based engines for non-dictation situations, they will generally have better results.
Cerowyn, I’m trying hard to believe you, but I just tried the MS speech-to-text program and it did not understand a single word I said. Not one. “The quick brown fox jumped over the lazy dog” was transcribed as “Law or OM were all were even.” It’s not even coming up with words. Absolutely nothing about the program worked. OTOH, saying the same phrase on my cell phone came up with the correct phrase.
So, I’m kind of at a loss as to what I might be doing wrong with the MS program.
Did you run the training wizard? Dictation engines usually need a good voice model, and the training wizard will also identify any audio issues that might play havoc with recognition. If you did, I don’t have an explanation for your experience. I don’t use the SR engine very often, but I just tried the same sentence as you in WordPad and it came out fine.
Here’s an Ars Technica review that judges MS’s effort good (but inferior to Dragon Naturally Speaking 9).
It doesn’t appear that Dragon Naturally Speaking has a demo version, but if you’re unhappy with the built-in SR engine, it might be worth the $99 for you to try out DNS.
I should also point out that my post was a bit inaccurate in at least one respect: Microsoft does have a server-based product, based on technology they bought in 2007.
I have both Dragon Naturally Speaking and Vista Voice Recognition. Both performed about the same for me. And I used the moduals to get the programs to adapt to your voice. Maybe I just speak awful