An unofficial blog that watches Google's attempts to move your operating system online since 2005. Not affiliated with Google.

Send your tips to gostips@gmail.com.

February 16, 2007

Testing Vista's Speech Recognition

A very patient individual decided to test Vista's speech recognition. Admittedly, it's not an easy task to build a software that actually works well and Microsoft doesn't have a brilliant history in this area (remember the demo from July last year?).

Microsoft is proud: "Windows Vista includes a built-in speech recognition user interface designed specifically for users who need to control Windows and enter text without using a keyboard or mouse. There is also a state-of-the-art general purpose speech recognition engine. Not only is this an extremely accurate engine, but it's also available in a variety of languages."

But this long video shows how difficult is to browse the Internet if you're brave enough to use only your voice, and not your fingers. To improve the performance of speech recognition, users should perform some training before actually using it. But it's weird to see that only Internet Explorer has problems, while Firefox is responsive.


This reminds me of a Saturday Night Live sketch (transcript), that illustrates how hard speech recognition is.

33 comments:

  1. I love how Firefox worked a lot better than IE. And it looked like when it was working that the speech software worked really well, not faster than typing or a mouse but really good none the less

    ReplyDelete
  2. Oh, just screw me...

    LOLLOOLLOL

    Speech recognition is still a long way from reality. Computers are going to have to 'think' as a human before this can truly come to fruition.

    ReplyDelete
  3. I have been using NaturallySpeaking 9 Preferred with windows XP and I'm quite happy with it (in fact, I am even using it while I'm writing - or shall I say speaking - this post).

    When I heard that Vista has a speech recognition, I didn't think of it being any more than a child play.

    Maybe Microsoft should stop putting out degraded software (just to put out any kind of software) and focus on developing quality products.

    ReplyDelete
  4. To be perfectly honest, I had great difficulty in understanding some parts (!) so it's no wander a PC does. That voice was filled with "urmms" and very quick speech.. I am not defending Microsoft, just voice recognition in general. I have Dragon Naturally speaking - version 7 - works a dream with a little practise of removing our involuntary noises. I imagine later versions - as Carl has suggested - work even better. As a side note, i am sure that if Microsoft brought out a comparable product bundled with any version of Windows, Naturally Speaking owners and others would be rushing to court like the AV providers etc etc, not to mention the EU, would they not? The inbuilt "solution" is, to me (and should be carefully marketed as such), merely a useful tool that can help out navigating around. It shouldn’t be compared to a full blown product that, incidentally, can cost the same, or more than all of Windows put together..

    ReplyDelete
  5. I'm totally agree with William Dowell ... this a junk.... You can't speak to a computer like that ad expect to get a good result... I've tested Vista's speech recognition without training it first and I've got good result but with a little training it works very smooth. I'm totally defending Microsoft because I'm a programmer and I know how hard it's to do simple thins, simple for other people that don't have the basics of programming. What Microsoft tries to do it's brilliant.. think about Linux it comes with all that stuffs but there are hard to use for first time users, Microsoft tries to simplify as much as it can and from my point of view it has a big success. Good job Microsoft, keep it going.

    ReplyDelete
  6. Speech recognition is supposed to be a tool for people who are handicapped in some way. I find it amazing that this "dude" is still able to open the cryptic url www.ytmnd.com with little problem and click on the "sexy chick whatever" link, which he said remarkably fast. people who have no real use for this system will of course try hard to break it. and who doesn't love a bit of corporation bashing?

    ReplyDelete
  7. Wow, does the thing not recognize the letter 'n'? Jeez, that's horrible.

    ReplyDelete
  8. You really can't do additional training for the Vista Speech Recognition. There are two (very boring) training scripts, and you just alternate between the two if you try to do more training.

    Also of note is the fact that you can't dictate into, say, Gmail in Firefox with Vista Speech Recogntion, although it works in IE. NaturallySpeaking has worked with FireFox since version 8 or earlier.

    I would use NaturallySpeaking, but version 9 refuses to install on Vista.

    ReplyDelete
  9. If you're already trained in Dragon NaturallySpeaking 8.1 or 9, the transition to Vista's SR is a breeze. You can't expect it to come instantly; training the software and training yourself are both essential. The additional "speech training" comes as you voice-correct the software in normal use. As for dictating into gmail , for example, go to speech recognition options and choose Enable Dictation Everywhere (I just tested it to make sure it would work). I think Microsoft's latest effort in speech recognition technology is inspired and such a far cry from its earlier efforts.By the way, I am dictating this comment using Vista speech recognition.

    ReplyDelete
  10. Can no one see that this guy doesn't really know what he is doing?
    To get it to push 'n' all you have to say is (and this is if it doesn't recognize you right the first time) "Press N as is nick"
    And I think the address bar in IE is just address. I'd test this, but I can't find it on my system.. WOOT! lol.
    Firefox isn't all that great at it but I think its just more the web content. If you watch the little voice rec. bar at the top it understands most of what he is saying, 'cept for the n thing.
    Also for pressing keys you can do more than one at a time. I started using Voice Recognition last night for abt an hour and I can do better than that. Plus, mousegrid rocks.
    I don't like the fact that you have to use the mouse for UAC though.. doesn't seem right.
    Later,
    Nick

    ReplyDelete
  11. Looking at the spelling and grammar, I'm guessing half of the people posting here need to use voice recognition because of illiteracy. Independent reviews of the Vista voice recognition software say it falls just short of the latest version of Dragon NaturallySpeaking. Not too bad for a "free" program. The biggest problem with Vista is it has too much bloat (like VR software) and consumes way too much memory. Most of the midrange laptops being sold with Vista INSTALLED do not have enough RAM to function properly using Vista...... WHAT A JOKE!

    ReplyDelete
  12. In defense of speech recognition, that computer is really LOUD! I don't any speech recognition software would have been able to understand that guys impatient uneven chatter through that noise without at least some training.

    ReplyDelete
  13. I am using Dragon NaturallySpeaking 9.5 to write this in Vista. I haven't yet tried the endogenous speech recognition software, but haven't heard good things about it so far (David Pogue at the New York Times has tried it and says it's okay, while he loves DNS). For me, the hardest thing about dictating is knowing what you're going to say before you say it. Once you have that worked out, and speak clearly using a high quality microphone, things work pretty well.

    ReplyDelete
  14. 'n' and 'm' sound so much alike. so yes, you are right. you have to say something like "m as in mary" or "n as in nancy"

    ReplyDelete
  15. firefox works better cause ienternet explorer wants to dictate :P

    ReplyDelete
  16. Actually if you read the help file it tells you 'Go to Address' goes to Address Bar.

    And if you right click the icon on the taskbar, you can modify the dictionary (record a new word or how you say a word).

    Aaand I can't get it to open Firefox.

    ReplyDelete
  17. It used to open FireFox just fine for me, but I recently tried it and it won't work just by saying "Open FireFox."

    However, if you say "Double click FireFox," it works. Not sure why the other way stopped working magically.

    This guy in the video obviously doesn't know what he's doing. Why review something if you're not going to/can't use it properly? This speech recognition works great, don't be fooled by watching an idiot fumbling with it.

    He used the "Correct" command to tell it to turn what he said into a blank space, for instance. Anyone who's spent 5 minutes doing the tutorial knows you only should use the "Correct" command when the computer made a mistake, not when you make a mistake.

    Basically, watching an idiot spend 5 minutes trying to go to a website because he didn't do the tutorials and thus doesn't know the commands is very misleading. Anyone who wants to give it a FAIR chance, I recommend taking the tutorial, do like 5 minutes of training, and try it yourself. You'll realize how dumb this guy is. There's no way he took the tutorial, or he might actually know a few of the commands. Is he seriously telling it to "delete backspace?" *Headdesk* Everyone knows "scratch that" or "delete that" is how you remove what you just said.

    Sigh.

    ReplyDelete
  18. I use Vista speech recognition for all text input and computer controls.
    It works accurately so I never use the keyboard or mouse now.
    Those who ridicule it have probably got a cheapskate headset.

    ReplyDelete
  19. So what's NOT a cheapskate headset?

    ReplyDelete
  20. Something I found out quite by accident (not documented ) was that you could say "Start Typing". Until you say "Stop Typing", it lets you spell words without having to say "Press" for each key. Handy eh?

    Sven

    ReplyDelete
  21. I am currently learning to use vistas speech recognition ability And I am lovin' it.

    ReplyDelete
  22. Have been using SR for eight months due to health issues, the thing is driving me crazy. If I happen to have a bookmark at my IE which is named something like the command I am saying, it just might open that bookmark instead of say, opening a file...
    Sucks.

    ReplyDelete
  23. The speech recognition system is very handy; I have been using it because of health issues as The only problem is it doesn't recognize a lot of the windows that I used to type in as windows that can be used to type in, such as my e-mail system or Open Office Writer.

    ReplyDelete
  24. Vista's Speech Recognition has really surprised me...for a Built-in application it's really quite good. I've only given it a small amount of training and it copes well. Even the most expensive SR software you can buy can't cope with everything.
    Never though I would ever say this but there are some good programmers at MS.

    ReplyDelete
  25. Wow, that stressed me out just watching it. I remember how excited I was to try that feature after installing Vista -- and haven't used it since. I'm wondering if Dragon is any better, or if you have the same screwy results?!?

    ReplyDelete
  26. It works great on my pc, i'm useing a high quality headset mic.
    Its not excellent but you can still do every thing. I don't care if i have to repeat some things a few times!!!!

    ReplyDelete
  27. I just discovered Vista's SR technology and let me tell you I'm loving it. I have done two full training modules, in addition to the initial setup. Listen folks give it a try before you knock it or listen to DNS or Apple fan boys (btw I typing this on my iPad so get off my back about being a MS groupie). I'm getting, yes even with the famed "Rainbow Passage" (Google it you'll understand Phonemes and their relationship to both speech and SR software) an average of 115 wpm with 95-97% accuracy. Make sure you have a good headset mic, I use a Sennheiser, and take time to both train the software to your voice and to learn the commands unlike the horse's ass in this video. I was going to purchase DNS 10 but hell I'll save the 100 bucks for something else. Why buy something when you have an alternative that works just as well?

    ReplyDelete
  28. I'm with Florin on this one; as a programmer I've been amazed at what Vista's SR CAN do! This is a very high quality piece of software that's just as free as your Spider Solitaire and Calc! Like any good technology, you have to expect to take the time to really learn it and work with it before you can expect the best performance from it. I suspect that some people have trouble with SR for some of the same reasons they have trouble in life: They don't think about what they're saying before they say it and they don't think through all the consequences their words will have. Just like when I'm talking to another person, if I say something to Vista and it misinterprets, I don't scream or "uhh!" at it. Instead, I try to figure out what it thought I said and then clarify. You have to meet technology halfway; the best features of, say, Photoshop require at least some level of expertise on the part of the user. Luckily, SR is easier than Photoshop! Speech recognition is a NEW technology still. I'm excited to see where it will still go in my lifetime.

    ReplyDelete
  29. LOL!
    Press "N"
    BackSpace
    Press "N"
    BackSpace
    Press "N"
    BackSpace
    Press "N"
    BackSpace
    Press "N"
    BackSpace
    (X3)

    ReplyDelete
  30. I just tried Vista VR, and it is a piece of crap with a 15$ desktop mic. Don't even bother experimenting unless you've got a top-of-the-range headset mic.

    The funny thing is, on skype and when recording voice messages, I would say my mic has very clear, decent sound quality. So I don't understand why the VR wouldn't be able to function properly with it.

    ReplyDelete
  31. Dude really??? You talk to fast. You don't enunciate and you sound like you have something in your mouth. I use speech recognition all the time and it is 98% flawless. You gotta learn how to talk first.

    Dude if you say uuuuuuuuuuu N then you are going to mess it up. Just say N if you mean N

    ReplyDelete
  32. I just started using Voice Recognition on Windows7 today. I did all the training etc. And bought a pretty good headset from Radio Shack. What I can't figure out is how to use it while using the internet. I say "address bar" and a couple of other things to enter a web site but no luck. Also my G Mail account, I can't figure out. And changing between web pages when you have more than one open. I run IE, Mozilla and Flock.So unless I can figure how to do all this. I will give up and return the headset. It sounded like a great thing until you dive into it. All commands have to be precise in what I am seeing. Any suggestions would be helpful. Tom

    ReplyDelete