Windows Vista and Speech Recognition

admin's picture

It appears the group's conjecture on MS and SR in the next windows release was pretty accurate. How well it will work only the beta testers can tell - if they can.

I found this blog with quite a bit of information on the new combo. It appears he may be involved in it at a developer level.

Rob Chambers - http://blogs.msdn.com/robch/

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
ScottW's picture

MS SR Beta Video

A few days ago I set up a wiki page to keep track of the goings-on about the new Microsoft speech recognition system.
I finally got the beta working, and uploaded a video.

See it  at the Microsoft speech recognition watch

 

-- Scott W
more Speech resources at SpeechWiki
Products at Applied Recognition

Rob Chambers's picture

Accuracy and correction options

Glad to see that you got it working, Scott. Thanks for uploading the video so people can see some of the UI and the interactions.

In the video you commented that the accuracy wasn't that great. That's something that the core engine team is continuing to work on, and will continue to work on throughout the rest of the Vista ship cycle. Acoustic models and language models will continue to be tweaked right up until the very end.

Also, the SR engine that's included in Beta 1 is almost 6 months old now. I'm pretty sure you'll see a difference in accuracy once we release the final version.

You also mentioned in the video that the correction options were limited. That's definitely true in the beta.

I wish I could talk publically about what we're doing to make that better for Beta 2, but since Beta 2 isn't out yet, I can't. Smiling I can say that you'll be happy with the changes we have already made in current Beta 2 builds for correction.

But just to make sure, what types of additional correction facilities would you (and would anyone else) like to see?

--robch
Architect
Windows Speech Recognition
http://blogs.msdn.com/robch/default.aspx

Windows Vista and Speech Recognition

Although I have not tried MS Speech as yet, I have been following the MS Speech Group, and have watched the development with interest.

My only suggestion is to take the best of ViaVoice 10.5 and DNS Pro 8, put them together, and you will have as near to a perfect product as you can get.

I would venture to suggest, as I always have, that ViaVoice is the more user friendly of the two. In my view (and I will be contradicted by the DNS users), it offers greater options for saving words and phrases, a very easy to use correction window, easier macro creating, you can choose the programmes that you want to dictate directly into, makes the deletion of unanted words from the custom vocabulary easy, is more flexible on the position, size and location of the Speecbar and most of the glitches have been ironed out. From an accuracy point of view, the there is very little to differentiate betwee them.

I would presume that the MS Speech development includes being able to dictate into, and command, any programme.

As regards navigation macros, for example in ViaVoice, I made macro where I simply say " speech computing", and this opens Mozzilla Firefox, opens Speech Computing, and goes straight to Recent Posts. Creating this macro took approximately one minute including fine-tuning the timing.

Finally, please bring out the UK English version.
Quentin

Windows Vista and Speech Recognition

Or even please, please, please bring out a UK English version.
Graham Hendry
www.itspeaking.co.uk

Microsoft Speech requests

I have two aspects that I am hoping for with the upcoming Microsoft Speech.

First, a Medical version, comparable to Dragon Medical 8.0. Hopefully accuracy will be as good with as short a learning curve.

Second, tight integration of Speech with the Tablet Input Panel. I hope it will be very easy and natural to dictate on a slate Tablet, using the pen to turn on and off the microphone, and using the pen to make corrections. Hopefully as one adds new words to the Tablet dictionary via the pen, one will be able to train the word for speech as well, easily and intuitively.

I currently use my Toshiba tablet using Dragon for speech and the TIP for making corrections, which works well but is not tightly integrated at this time. I'm greatly looking forward to Vista, hoping that this will continue to advance these features.

Joe

Speed of Microsoft product

Scott,

I'd be curious to know how the speed of MS speech recognition versus the Dragon. (I recognize that this is a Beta and a lot could change.)

Matt Chambers (no relation to Rob Chambers, as far as I know)

ScottW's picture

responsiveness

I do not. My subjective impression was that the responsiveness of Microsoft speech was faster then DragonNaturallySpeaking when using a medium-size vocabulary (I don't know how large the Microsoft vocabulary is).

The demonstration was done on a virtual machine, so it's reasonable to conclude that it would be faster if it was running directly on hardware.

One other thing is that the Microsoft speech product does not have anything like the Dragon results box which gives users a visual clue as to how long recognition is taking, so even when it's very fast, because you can see the progress you still get a sense that time is elapsing.

Another thing I noticed recently is that the Microsoft speech engine (still) doesn't seem to have any dictation support for nonstandard Windows

 

--
Scott W
Speech resources at SpeechWiki
Products at Applied Recognition

Rob Chambers's picture

Non-standard windows

It's true. Currently, Windows Vista Beta 1 Speech Recognition doesn't have support for non-standard "Edit" type controls.

Windows SR is using the Text Services Framework directly to speech enable "Edit" type controls. If the control doesn't support TSF, we're not speech enabling it, currently.

What Controls and Appliations would you like us to work with in addition to those listed in my blog? Let me know. I'd love to hear your opinion.

--robch

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.




view recent posts