Dialog Managers

Flat
up-to-date review of ready-to-use linux speech recognition front-ends
User: JohanLingen
Date: 12/22/2007 6:40 am
Views: 25592
Rating: 38

Hello all,

 I read the information on this website with much interest. I'd like to install Kubuntu for a spastic friend, but she needs some kind of speech recognition software. Right now she is kind of forced to use Windows because that's the platform that Dragon Naturally Speaking supports.

Can anyone provide a current overview of easy-to-install  front-ends for linux?

 And of course I will now submit my voice example :-)

--- (Edited on 12/22/2007 6:40 am [GMT-0600] by JohanLingen) ---

Simon project
User: ralfherzog
Date: 12/22/2007 12:17 pm
Views: 396
Rating: 66
Hello,

You should take a look at the Simon project (obviously only in the German language available).

In my opinion, Dragon NaturallySpeaking 9.5 is the best solution.  For example, you can say "MouseGrid", and navigate the mouse by using the voice.  This is extremely comfortable, and I am convinced that we will need several years to develop an equivalent open-source solution for a system like (K)Ubuntu.

Your friend should continue to use Windows.  It is probably the best system.

--- (Edited on 2007-12-22 12:17 pm [GMT-0600] by ralfherzog) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: kmaclean
Date: 12/22/2007 2:56 pm
Views: 441
Rating: 32

Hi JohanLingen,

>Can anyone provide a current overview of easy-to-install  front-ends for linux?

You might want to look at these command and control speech recognition front-ends (from the VoxForgeDevWiki):

>Right now she is kind of forced to use Windows because that's the platform

>that Dragon Naturally Speaking supports.

For dictation on Linux, the OSSRI group seems to have had some success installing DNS9 on Wine, there was something in their mailing list, but I can't seem to find a link to their archives - you might subscribe to their mailing list and ask them or look at the Wine project site.

Ken 

 

--- (Edited on 12/22/2007 3:56 pm [GMT-0500] by kmaclean) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: JohanLingen
Date: 12/24/2007 4:59 am
Views: 338
Rating: 29

Thank you both for replying, I hadn't heard yet of Perlbox and Speechlion. I can only see Gnome-voice-control in the repositories. Does that mean that the other programs are not ready for use yet?

For now I tried Perlbox, the installation manual was not that hard, so I got it running. The only think I can get it to do right now, however, is starting or focusing to kmail, konqueror or kate. "Console" gives me konqueror again. Is this program also able to type the text as I speak it?

 I've looket at Dragon on Wine, but it came with an extensive manual and the end result would be a partially functioning application. Maybe you meant just this link: http://appdb.winehq.org/appview.php?iAppId=2077

 

--- (Edited on 12/24/2007 4:59 am [GMT-0600] by JohanLingen) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: kmaclean
Date: 12/24/2007 7:43 am
Views: 374
Rating: 26

>I can only see Gnome-voice-control in the repositories. Does that mean that the other programs are not ready for use yet?

Not necessarily ...  though it is better to stick with something included with your Linux distribution.

>Is this program also able to type the text as I speak it?

No. 

There is no Open Source Speech Recognition Engine for Linux that permits you to "dictate" text into your microphone and have it display on your screen.  We need better acoustic models.  To get better acoustic models, we need lots of speech - please donate your speech to VoxForge.

Open Source Speech Recognition on Linux is currently limited to "command and control" type applications - where you can use your voice to open windows, switch programs, and select menu items using your voice.  This is not the same as a "dictation" application.

>Maybe you meant just this link: http://appdb.winehq.org/appview.php?iAppId=2077

No.  But you might email Susan Craig - she's with OSSRI and I believe she was the originator of the thread I was thinking of.

Ken

--- (Edited on 12/24/2007 8:43 am [GMT-0500] by kmaclean) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: JohanLingen
Date: 12/29/2007 4:06 pm
Views: 368
Rating: 34

ah, I was indeed looking for dictation applications :-)  Thanks for your information!

Too bad, there's no such yet, but great to hear I can actually do something to help it become available!

I was somewhat puzzled by the manual to hand in audio samples, but I think I can do it now. Now just wait until that cold I caught is gone...

 By the way: I gave up on gnome-voice-control, the daemon wouldn't start listening and I could not find documentation on the commands I could issue. Perlbox stays and I hope to try Simon in the future (when it comes in a package).

 Best regards,

 Johan

--- (Edited on 12/29/2007 4:06 pm [GMT-0600] by JohanLingen) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: kmaclean
Date: 12/29/2007 5:37 pm
Views: 365
Rating: 42

Hi Johan,

>I was somewhat puzzled by the manual to hand in audio samples, but I think I can do it now.

How can we make things easier/clearer for others to submit audio?  Any feedback would be greatly appreciated.

thanks,

Ken 

--- (Edited on 12/29/2007 6:37 pm [GMT-0500] by kmaclean) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: DavidGelbart
Date: 12/31/2007 3:30 pm
Views: 349
Rating: 35

An alternative to running Dragon on Wine would be to run Windows on a virtual machine (or to run Kubuntu on a virtual machine under Windows)  and use WinDictator.

http://foss.eepatents.com/trac/WinDictator/wiki 

I'm not a Dragon user, but I suppose Dragon running under Windows integrates with applications (and I guess the operating system as well) in various ways.  I suppose WinDictator only supports a subset of this. So I suppose sticking with Windows is the best option for now, unless she has a strong interest in running a Linux desktop.

 

 

--- (Edited on 12/31/2007 3:30 pm [GMT-0600] by DavidGelbart) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: JohanLingen
Date: 1/1/2008 7:05 am
Views: 392
Rating: 35

The interest in linux is to avoid all the viruses and other malware that flood her computer as well as avoiding license costs, but I guess for now it's best to stick with Windows and reconsider in the future.

 If anything comes to my mind to improve the voice-submitting, I'll post it here (or elsewhere if you want). It wasn't that hard after all to record a voice, but it took me quite some time and effort to find out how exactly to do it (I used Audacity, because I didn't manage to adjust the microphone volume under KDE (using Kmix, maybe add?) ).  I will ask my girlfriend to follow the instructions, I think that might give some new ideas :-)

The Audacity manual is quite OK, but deviates at some points from my installation (under Kubuntu). There is i.e. no tab to select the output file format to .WAV (this can be done later, while exporting). Also, one needs to make sure that the play-back option (play one channel, while recording another) is disabled. Else, an error pops up (though this is due to my lack of understanding Audacity :-) ).

Also, I saw that some portions of the website are translated to dutch, and others are not. For dutch models, I think it is very helpful to translate all of the website and make the choice between English/Dutch more accesible (or even automatic). Then, you could also integrate the dutch voice files with the submit-process (as well as European English). I thought I saw them only in a forum topic now (still have to look at that).

 Looking backwards, I think it was quite easy to submit voice, but while I was doing it, I had the impression that I needed to collect some information from various websites myself.

--- (Edited on 1/1/2008 7:05 am [GMT-0600] by JohanLingen) ---

Re: up-to-date review of ready-to-use linux speech recognition front-ends
User: kmaclean
Date: 1/13/2008 1:45 pm
Views: 465
Rating: 31

Hi JohanLingen,

Sorry for the delay in getting back to you on this. 

> I didn't manage to adjust the microphone volume under KDE (using Kmix, maybe add?) 

Good idea .. .will do.

>Then, you could also integrate the dutch voice files with the submit-process

>(as well as European English). I thought I saw them only in a forum topic now

>(still have to look at that).

We are still working on a multilingual version of the VoxForge website - it is still in the very early stages.  However, the Speech Submission application is now available in Dutch.

Ken 

--- (Edited on 1/13/2008 2:45 pm [GMT-0500] by kmaclean) ---

PreviousNext