I configured Voice Recognition on Vista and then used it for most of my Parallel Architecture homework. Despite my only using minimal training instead of full training, it's a lot more accurate than XP recognition was; this is actually really usable, while XP required so much correction it wasn't worth the trouble. I really like not having to switch between Dictation and Command modes, although I'll have to look up a workaround for when what I want to write would be otherwise recognized as a command. (Although Vista figures it out in most cases by my tone of voice, which is a remarkably clever bit of code. I've tested this- say a command word, but change the inflection, and it will change whether it dictates it (in Word) or performs the command.)
Unfortunately, I've gotten slightly too used to its voice recognition, in a very short amount of time. At least when I'm in that mindset. Voice recognition is really convenient for me with how my computers are set up; it's not that convenient to type on my Tablet PC given how my desktop takes up my entire desk space (with its two monitors), so I probably actually will be using it; it's convenient to have homework in one computer and references and the assignment in the other. (And by keeping my homework on my Tablet, I take it with me to class- if I forget to print out an assignment, I can switch the most recent copy to a jump drive and print it off a lab computer with relatively little trouble.) But getting used to voice recognition can be a problem. Let's go back a few hours to when I decided I wanted a hard copy of the general voice command reference.
"What can I say?"
ding!
"Maximize that."
ding!
"Show all."
ding!
"Print."
doot?
"1." (It wasn't sure whether I meant Print or Print Preview, so it labeled the buttons with numbers to ask for clarification.)
ding!
"OK."(Numbered commands have to be confirmed.)
ding!
"Print."
doot?
"2." (This was ambiguity between Print and Print Settings.)
ding!
"OK."
ding!
*whirrr... shoof, shoof, shoof* (well, what do you think a laser printer sounds like, if not whirr shoof shoof shoof?)
*rustle rustle*
"Staple that."
blat?
"Staple that."
blat?
"St- wait, never mind."
blat?
"Stop listening."
donk!
*kchunk*
Maybe I need to implement voice recognition on my stapler next.
- Start Semagic; click login; go to Subject; Dear Aunt, let's set so double...
kistaro- March 7th, 2007
silly
I find it mildly amusing that the one of us who does not seem to be a dragon is the one using speech recognition software by that name. Then again, I am easily amused.
I want one of the ooooold silk banners that Dragon used to give their VAR's in the 90's. Oh my those were pretty. Pity their software sucked compared to VoiceType Dictation/2 at the time. IBM just owned that market utterly.
It definitely didn't help that I had a speech impediment. Most people had trouble understanding me, so it's not a surprise that it was basically a lost cause for my computer. Apparently, VoiceType cannot adapt to someone who nasalizes the letter-sound "R". (If you try to figure out how to nasalize an R- if it feels like you're choking, you're on the right track. At the time- and it took years to really learn otherwise- I thought that was correct.)
And, of course, yay for dragons. But then, I'm biased... (*is one of those scary insane spiritualist people*)
That would be keen. Then when I get asked by 328420394 students every day do we have a stapler, or where is the stapler, it would just zoom up and present itself to them.
Implementing the hoverjets for the stapler so that it could in fact "zoom up" is left as an exercise for the reader.
Voice activation can be done by a computerized base station, although in the non-quiet environment it would likely be in, it would probably have too many false alarms to be practical. Individual call buttons at possible stapler-retrieval points would work better.
All things considered, hoverjets would probably be more practical.
...but darn it, don't little plastic robots have more style?
Ding!
*tap*
Ding!
*click*
*tap tippity tip tap... tap*
Ding!
I was going to try to dictate what printing something does, but it just goes *whrrrr* with no end in sight. Time to fix my printer...
I think the largest usability problems they have, from when I've used them before, are to do with punctuation, though. Do you have to say "Open quotes stop listening period close quotes" or the like? How about "It wasn't sure whether I meant Print capitalise that or Print capitalise that Preview capitalise that comma..."?
I also wonder if they localise for non-Americans...
Yes, punctuation is inconvenient. You are almost verbatim correct about the punctuation- except I added a few "literal" in there to make sure it didn't interpret my command as a command. ("Literal" is the escape, forcing it to interpret the next word as dictation; "literal comma" is required to get "comma" instead of ",".)
I know there's a setting that would understand you, although other localizations aren't too far along- other than a few variants of English, you've got Japanese and Chinese. Guess what countries the original researchers were from...
It does have a really clever feature to dramatically enhance recognition: if you let it, it will read all your e-mail and documents and look for words it doesn't know. It will tentatively add them to the dictionary as possible things to recognize, with unknown pronunciation or part-of-speech. This is how it got the words "draconity", "draconid", "Otherkin", and "motherfucker" right the first time I said each of them.
Obviously, it can also be explicitly taught new words- if it misrecognizes something, if it doesn't have it listed as an alternate- and saying the word again doesn't bring it up as an option- I can spell it, and if it still doesn't know the word, it will add it to the dictionary- and explicitly ask about capitalization and for a very clear reference pronunciation to maximize the chance of recognition.
Inconvenient exactly once, but then it tends to remain correct!