Wow, I am pleased to be the first person to be the first person to comment on using voice dictation with the tablet.

I have been struggling with voice recognition ever since Dragon NaturallySpeaking version number 3 -- whenever that was. I have always used it on a laptop, and later a tablet, simply because I like the portability and I absolutely hate typing.

"The gurus" said that major speech recognition improvement had been made between dragon version 7 and version 8. I did not find that to be so. It is true, the nature of my misrecognition's changed, but I thought I was making equally as many errors. (And I must tell you it is that in correcting errors that speech recognition becomes a bore -- very simply, you must correct your errors. Many of them will make your text absolutely unable to be understood by any reader -- sometimes, even including yourself. It is not like typos where you can sort of parse out what you had in mind -- some of these "speak-o's" render your text unintelligible. Trust me.)

Having said that, for me, version 9 has been spectacular. Somehow, Dragon has finally got attuned to my voice. I make very, very few errors (actually, I don't make the errors -- Dragon does!).

I love using a tablet PC in the slate mode. I love the interaction with the stylus -- even find all I use is the stylus as a mouse. This version 9 is really opening up the tablet possibilities for me -- and making me much more communicative. I am willing to dash off an e-mail that has more than just a few words that I scribble, struggling with the tablet input panel.

Also, I am using Dragon NaturallySpeaking preferred on Gateway M 275 convertible with the Pentium Centrino 1.4 GHz chip. This barely meets the specs that dragon version 9 is supposed to require. (To be fair, however, I have 2 GB of RAM.) I don't find any problem, for the most part, with dragon overloading my system.

I do have problems with MindManager -- things start to heat up when I use MindManager alone, and when I use them both at the same time things get hot fast. Still, the fan will only go on intermittently.

Parenthetically, not every program works well with dragon. Where you can really see this is when you make corrections -- sometimes the cursor does not seem to know where it is. Unfortunately, OneNote is one of these programs. But keep in mind, the recognition in this latest version is so accurate, but that is really a very minor annoyance. I use dragon to make inputs into OneNote without hesitation.

One thing I have noticed about this version of dragon is that it is much less processor intensive when the microphone is put into "sleep" mode. It used to be that my processor would run at about 13% capacity when the microphone was asleep. Now it barely runs at two to 3% capacity that is not much overhead -- certainly not for the convenience.

As I am writing this, I am salivating for one of the new Bluetooth microphones. I have to buck up my career each to install a Bluetooth USB dongle on my system. (Please note, that was a recognition error. I have decided to leave my recognition errors in this text and, that has been my first -- at least the first one that I caught! :-) what I said was: "I have to buck up my courage to install...")

The issue of microphones is very important. Dragon will come with a microphone -- but all of the gurus say that those are pretty much worthless. You do want to buy a good microphone -- there are vendors who specialize in microphones for speech recognition. You can find them on the websites and chat rooms. (The issue of sound cards is also quite critical -- even more so in laptops and tablets. The gurus recommend that laptop users get a USB pod to act as your sound card.)

Nuance Software has recommended Bluetooth headphones, but all of them have the extended arms so that the microphone goes at the corner of your mouth. Dragon depends on getting absolutely every sound and being able to differentiate between aspirated and unaspirated syllables for its recognition. A microphone on your cheek is simply not going to "hear" that.

If this helps, I saw a picture of one Dragon NaturallySpeaking guru who were the headband of the headset around his neck and propped the microphone up to the corner of his mouth. He knew where to put his head when he spoke so that the microphone would be in the proper place.

One of the microphones that Nuance software has approved is the one that Rob Bush way reviewed. Once I screw up my current age (my, my, Dragon does not like how I say cordage, no, courage -- at this does make my point about how you can make your text non-sensible) I will be more than happy to order the microphone through the Text Smith Corp. that are sponsors of your really fine blog.

As for me, I recommend Dragon NaturallySpeaking version 9 wholeheartedly. From what I understand about the speech recognition included in This to (no, Vista), Microsoft has a long way to go in making corrections easy to make. And, like I argued before, not being able to make corrections easily can drive you nuts. (I have this same problem with the tablet input panel -- it is just easier to start over than to try and make a correction. Microsoft has a lesson to learn.)

I recently started experimenting with voice recognition since I got my Bluetooth headset and posted the (boring) results of a test on my blog - http://sumocat.blogspot.com/2006/07/audio-my-new-bluetooth-headset.html - not worth listening to, but the transcript provides a quick and easy visual.

Anyway, I haven't used voice recog to full effect primarily because of my hardware (first gen slate, Bluetooth headset with wide area mic). Dictation is not good enough to be standalone and if I'm editing text, I might as well ink or type it. Actually, voice command works quite well, but I don't often use my tablet in a situation where it would help to yell commands at it, especially since I can't customize them. I would gladly tell it to "tabletize" if it could comply.

I've used voice recognition with my Q1 and it worked quite well... until I caught the 'flu. Apparently my voice turned mutant or something; I have to retrain it now.

I think the key is a good input source; XP Tablet's voice recognition works well (not spectacularly, but well enough) with the built-in mic array and with the Bluetooth headset I tried out. Indeed, roughly half of my blog-review of the Q1 was done via dictation.

Once I've retrained the system to my healthy voice again, I'll have to work on the voice commands; I'd started to make some headway on getting them to work, but really haven't given them a good test.

-- Steve

I just ran across this explanation of the Microsoft demonstration on one of the independent websites that support Dragon NaturallySpeaking. (The following is a quote...


I just heard from Rob Chambers, one of the leaders of the Microsoft Speech group. He posted a message on his personal blog explaining what happened at the demo. I can confirm that I had the exact same problem concerning the audio subsystem and gain with a previous beta version. The latest beta version of this has fixed that but apparently was not being used by the Vista Speech demonstrator. Click Here for a link to Rob Chambers' blog explanation.

--
Martin Markoe
See us at eMicrophones

I've dabbled a bit with Dragon version 8, mainly using it for medical-legal reports. I haven't worked up the courage yet to try it for clinical data entry on my EMR. I, too, noticed a significant improvement with accuracy going from the less expensive headsets to a premium one. The Philips SpeechMike Pro 6274 I purchased is a nice USB handheld mike, with a handy little trackball built into it.

Perhaps if version 9 is that much of an improvement, I might be inclined to try it for my electronic medical records.

Just to throw a slightly different perspective on to the whole voice recognition thing - although I can envisage some degree of practicality from voice recognition and would love to take advantage of it, this method of computer interaction is all but useless for me. I have a very slight stammer, which although becoming less and less noticable as I get older, is still omnipresent enough for the recognition software to often look at me blankly while shrugging its virtual shoulders. I think this problem ties in loosely with your 'environmental condition' comment. If the voicee (new word yay!) is incapable of effectively dictating to the Tablet whether that be because of disability, cold, drunkeness (not me honest), eating etc. then voice recognition becomes as instantly useless as it would in a noisy environment.

I have been using DNS version 8 for quite a bit and just loaded version 9 this morning. It is great and, in fact, runs fine on a Fuji P1510d with 512 ram. As stated previously, a good headset is a must but, with it, the dication works GREAT and has become one of my main methods for text input. No, it is not perfect but then again neither is my typing. For me it is invaluable to the way I work.