Perhaps it's me; maybe I'm doing something wrong.  DNS 8 Preferred seemed to be learning my voice, or I guess I should say, my dictation, much quicker and easier than DNS 9 Professional with UniVoice.  One word that I cannot dictate without going through some time-consuming correction is the title of the software: You know voice.  I am not correcting that title purposely.  Of course what I am saying is UniVoice, but no matter how I try I cannot get DNS to recognize that in this version.  It is always recognized as it is above, or a close variation thereof.  Of course it is primarily only in this forum that I use that word, yet in DNS 8 Preferred the software was recognizing that correctly after the second correction.  This version has yet to recognize it correctly.  Each time I use that word I have to select it, spell it, choose the correct corresponding number, then select it again so that I can choose the correct capitalization.  (I cannot find any instruction at all on how to capitalize letters in the middle of words).  Every time.  It is set up in my vocabulary correctly, and obviously there's nothing I can do to eliminate the mistaken phrase that usually gets typed.  I know this is not a great example, because that is not a word that is used in everyday conversation, and thus, everyday dictation.  However it is an example that is replicated with many other phrases.

When I started using DNS 8 Preferred again this past summer after having given up on it several months earlier, my recognition accuracy was not good at all.  That is until I purchased this ME-3, and my accuracy improved tremendously.  However now, using DNS 9 Pro with UniVoice (by the way, I typed that with my keyboard -- I'm sick and tired of going through that same correction sequence over and over and over again!), my recognition accuracy is so-so, but it is discouraging that it does not seem to be improving at anywhere near the pace that it did with DNS 8 Preferred.  Additionally, adding that word, along with other phrases, to my vocabulary does not seem to increase the probability that DNS will use that when recognizing my dictation instead of grabbing random words that remotely sound like the phrase and inserting that instead.

If I sound frustrated is because I am!  Isn't everyone who is having SR problems?  Don't worry, I am not anywhere close to throwing in the towel; I have too much invested, and I am too committed this time around to give up - my disabilities won't let me.  All that notwithstanding, my recognition accuracy had gotten to an acceptable level with DNS 8 Preferred.  While I did not expect miracles by upgrading, I definitely did not expect such a surprising drop-back in performance.  That's why in the first sentence of this post I stated it might be me doing something wrong; with me, that's certainly not outside the realm of possibility!  Do any other UniVoice users here know enough about the software to tell me if I need to be adding phrases in a different way?

How do I train UniVoice?  Instructions indicate that I should not train this in the same manner that I trained DNS without this add-on.  "Skip the following X number of steps, because UniVoice does not need them", is used throughout the setup instructions.   How then should I train it?  The instructions for UniVoice are fairly sparse; after all in actuality it is simply a specially trained user file as I understand it.  A number of posts in this forum are from users who are having a problem with UniVoice, and a lot of times the answer is, “you didn't have to do that” or “you're not supposed to do that in UniVoice”.  I really don't want to learn here what I'm not supposed to do; I would much prefer good advice on what to do that is right with this software.  Does anybody here currently using UniVoice know?  Tips and hints based on your experience are really, really welcome!

I really appreciate any input on this. 

BTW, paragraph changes are not recognized by this forum software when dictating to it (Which is really not good for a forum dedicated to speech recognition!!)  I remember that Lunis once mentioned that there is a workaround he uses to force paragraphs to be recognized when posting using dictation, but I cannot find that reference anymore - does anyone know how to do that?  Thanks! 

-------------------------

J-Mac

 

64 X2 Dual Core Processor 4400+ ; Corsair Twin2X 2048 MB (2 GB) High-Speed DDR PC3200 RAM, non-ECC ; NVIDIA GeForce 7900 GTX ; Windows XP Pro ; Senheisser ME-3 with Andrea USB Sound Pod

The only thing different about UniVoice (which is due to be updated within the next 24-48 hours) is that we recommend against running the Acoustic Optimizer.  Other than that, you treat it exactly like any other trained speech recognition file.  You make corrections in the usual manner, add vocabulary, phrases and corrections.  Don't forget to save your user files ;-).  We have never experienced difficulty with dictating “UniVoice”, other than correcting it the first time but no 2 voices are alike.  Although you can always train a word or phrase, it is time-consuming and we would only recommend it as a last resort.  We recommend utilizing your Vocabulary Editor’s Written Form/Spoken Form with the Spoken Form being “UniVoice” and the Written Form being “you know voice”.

Note: When we first attempted to dictate “you know voice” we got “you know boys”.  We often experienced difficulty with NaturallySpeaking differentiating “voice” from “boys”.  As mentioned previously, each end-user will experience unique speech recognition problems because of the uniqueness of their voice.

-------------------------

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012 

 

 

 

Hi Lunis.

Updating it?  That sounds good!  I would imagine that has something to do with support for the Preferred edition.  Are there any new features or other improvements for DNS 9 Pro?  

I hear what you're saying, Lunis, but why would I find it more difficult with this version than with DNS 8 Preferred? I am using the same PC, at the same desk and in the same room that I was using DNS 8 Preferred up until a few days ago.  Yet the difference in recognition accuracy is like night and day.  The only differences are the version of DNS that I am using, and the user file.  And DNS 8 Preferred the user file that I had created and trained was probably done in as amateurish a fashion as could be.  Yet my accuracy with that user was much better than what I am seeing now.  I don't think that my voice has changed that much in a few days, and I am speaking the same way as I did then.  Although I will admit that I find myself slowing down and enunciating each word more slowly, mainly because when I speak at the same pace as I used to the recognition accuracy is horrible.  In this post alone I am making two to three corrections per sentence!  That is worse than the first time I used DNS. 

All else seems to be equal; I cannot determine anything different in my dictation today compared with a few days ago, other than the software.  If I had not experienced this first hand I don't know if I would believe it myself.  I have to admit, it is a little disheartening.  While I know that you will work with me to get this right, a full day or two of making two to three corrections per sentence can wear anyone down.  

Now that's the kind of tip I was hoping to see!  I have never worked with Written Form and Spoken Form before, so this is a big help in itself.  I have read the Dragon NaturallySpeaking manual to the point where I almost have it memorized, yet I know that there is instruction there that is often not recommended by many in this forum.  After a while I become very gun shy, unwilling to perform much of the training or other actions directed in the Nuance manual for fear of doing something counterproductive.  The Acoustic Optimizer is a good example -- the manual strongly recommends it but I knew not to try something like that without checking here first.  The problem is I'm starting to feel that way about most things in the manual, wanting some confirmation before I try any of it.  I feel as if I've taken several steps backward with this software upgrade.  I know that isn't true, but what I am suddenly making this many more corrections it sure feels that way!

A few quick questions:

1. The recognition delay, that is the delay between me speaking and DNS inserting the text, is much, much longer than it was before I upgraded.  Yet the speed versus accuracy slider is on the second tick from the left.  Is DNS 9 Pro that much slower?  Or should I move that slider?
   
   
2. In DNS 8 Preferred the Dictation Box was on top unless I clicked my cursor outside the box.  Then the Dictation Box would act like any other Windows dialog box and move behind whatever other program I clicked on -- usually the program I am dictating into.  Now, however, the Dictation Box always stays on top and I have no access to the window behind it.  This is inconvenient because if I need to look at something in the window behind the Dictation Box I actually have to close the Dictation Box first, check what I need to, and then open it again.  For example, if I need to look at the post I am replying to, or if I need to copy part of it for a quote, I can't! Why has this behavior changed?  Can I change it back?  Using the Options button in the Dictation Box, I changed it from the Fixed to Cursor location, but that did not help.
   
   
3. How can I make the forum software see my paragraph breaks using dictation???

Thanks Lunis.   

-------------------------

J-Mac

 

64 X2 Dual Core Processor 4400+ ; Corsair Twin2X 2048 MB (2 GB) High-Speed DDR PC3200 RAM, non-ECC ; NVIDIA GeForce 7900 GTX ; Windows XP Pro ; Senheisser ME-3 with Andrea USB Sound Pod

Do not feel alone, J-Mac,

I have corrected “you know voice” probably 50 times.  I cannot get “you know voice” to dictate “you know voice” even after training it probably 10 times.  Hopefully, within the next 24-48 hours I will be able to do so.

I can tell you that your frustration level will quickly go down because of the help you get on this forum.  Over the past few years since I discovered KnowBrainer voice recognition has become much easier for me because of all the help Lunis and others give.  For example, Stu Miller help me create a couple of special commands which have been invaluable to me.

Look at the forum as the old “cavalry” coming to the rescue.  The only unsolved problem to date is how to dictate the “you know voice” and oddly with “you know voice” it does not usually get “Lunas” correct.  In fact, that is what I always get when I dictate the name.

Take care.

Jim Triona


-------------------------
Triona

You mentioned that you trained “you know voice” a number of times but it didn't stick.  This is an obvious case where training is the wrong approach.  As we stated many times in the past, training should only be used as a last resort. You need to find out what NaturallySpeaking wants to type when you say the phrase and use that phrase against it.  One of the following 3 options should be the magic pill.

Option #1.  The simplest solution is to add a commonly used word into a phrase and add that phrase into the Vocabulary Editor.  Example: We could never get NaturallySpeaking to learn the word “pod” so we added it into a phrase and added “USB Pod” to our vocabulary.  Because “UniVoice” is a single word, option #1 won't help.

Option #2.  Written Form/Spoken Form.  As previously mentioned, you could add “UniVoice” to the Written Form and “you know voice” to the Spoken Form.

Option #3.  Type “you know voice” into the Written Form, click the Properties button, put a checkmark in Use alternate Written Form and type “UniVoice” in the field.

-------------------------

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012 

 

 

 

Thank you for your response, Lunis,

I never thought of the Vocabulary Editor.  I was having trouble with your name which was always coming out “Lunas.” so that appears fixed.

Now I'm going to say that difficult word for the first time: “UniVoice” -- oh happy day it worked.

By the way, is UniVoice 9.1 going to cost us anything?  

Jim

-------------------------
Triona

I just added UniVoice in Written Form and "you know voice" in Spoken Form.  It did work, but not right away.  As soon as I set it in the previous sentence it came out wrong again.  Pulling up the Correction Box did not show the correct Written Form either.  I had to delete the entire sentence and re-dictate it three times before it was correct.  This just isn't working well. 


-------------------------

J-Mac

 

64 X2 Dual Core Processor 4400+ ; Corsair Twin2X 2048 MB (2 GB) High-Speed DDR PC3200 RAM, non-ECC ; NVIDIA GeForce 7900 GTX ; Windows XP Pro ; Senheisser ME-3 with Andrea USB Sound Pod

Lunis,

After my last post I did some checking and to my surprise I found the file ctfmon.exe listed as a process in the Windows task manager.  I remember reading about that file being related to the Microsoft speech engine.  I looked at my startup programs and processes and that file is not listed there, so it is most likely being called by another program.  If Microsoft speech has been enabled on this PC, I am not certain how it got that way.  Perhaps it was enabled by the manufacturer for one of the pre-installed applications.  I would imagine that could be the cause for some of the difficulty I am having.  I need to read my manual again and follow the steps to disable or remove that.

Any advice before I get started?

-------------------------

J-Mac

 

64 X2 Dual Core Processor 4400+ ; Corsair Twin2X 2048 MB (2 GB) High-Speed DDR PC3200 RAM, non-ECC ; NVIDIA GeForce 7900 GTX ; Windows XP Pro ; Senheisser ME-3 with Andrea USB Sound Pod

Just so you don't feel lonely, J-Mac,

I have that process on my task manager so I'm interested in what the great man, Lunis, has for advice. As I recall, Windows XP had its own embedded speech recognition program. I don't think anyone has ever said anything about shutting it off or eliminating it.  

 I used his option number three to fix my UniVoice problem.  And it appears to be fixed.

Keep it up.  I am sort of tagging along with you today.

Jim

-------------------------
Triona

this reference was posted earlier and answers FAQ's for Ctfmon.exe.

http://support.microsoft.com/kb/q282599

there is also something on the Nuance pages

-------------------------
Jomark

Okay, all Microsoft speech components are gone.  Vamoose.  Kaput!

Startup configuration, registry edited, Microsoft Office installation changed.  PC rebooted three separate times.  So there is no longer any Microsoft speech component that can pose an obstacle to perfect recognition accuracy...  Well, okay, maybe not perfect, but at least as close as reasonably possible!

Jim -- Here is the link that I found that leads to a Nuance knowledgebase article which explains how to remove all traces of the Microsoft speech engine.  The largest part of that is usually installed with Microsoft office.  I found this by performing a search of the forum, which led to a thread where someone asked a similar question about removing the Microsoft speech engine.  User Graham provided this link in his reply to that post. 

http://knowledgebase.nuance.com/view.asp?tnID=3244&sQuery=3118

Hey Lunis -- I just had to close the Dictation Box in order to copy the above link so that I could paste it here for Jim.  I did not have to do that previously, in the last version.  If you know, please tell me what to change to restore the Dictation Box to that previous behavior.  Thanks!

@Jim:

The instructions in the above article involve some semi--serious tinkering in Windows; be careful, and make sure you back everything up before doing anything.  I don't mean your Dragon files, but your Windows system files and registry.  

-------------------------

J-Mac

 

64 X2 Dual Core Processor 4400+ ; Corsair Twin2X 2048 MB (2 GB) High-Speed DDR PC3200 RAM, non-ECC ; NVIDIA GeForce 7900 GTX ; Windows XP Pro ; Senheisser ME-3 with Andrea USB Sound Pod

Thank you so much, J-Mac,Now to show you how limited I am, how do I go about saving the Microsoft registry and files as you suggest?

As usual I didn't read far enough. I went to the Nuance link and it clearly showed me what to do. I have the registry on my desktop. 

Has your accuracy improved since you removed all of the Microsoft speech engine?

I'm going to do a separate reply to Lunis to see if he agrees with doing so. He is the master of this universe!

Appreciate your help so much.

Jim

-------------------------
Triona

I have to have the master's input!


Do you agree with J-Mac about removing all the references to the Microsoft speech engine?

Thank you so much.

One of the disciples,

Jim Triona

-------------------------
Triona

It certainly doesn't hurt

-------------------------

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012 

 

 

 

That was some kind of ringing endorsement, Lunis,

I was hoping you would say it is the magic bullet to give us 101% accuracy.

Have you gone to “spin” school?  If I ask you your favorite color, are you going to tell me “plaid?”

Seriously, so I guess there is no problem removing it.  

As always, we wait with unbridled anticipation for your pronouncements.



-------------------------
Triona

Jim,

It certainly cannot hurt to remove Microsoft Speech. However, I have been using Dragon NaturallySpeaking and Microsoft Speech for Office 2003 on the same computers for over three years. As far as I can tell, there has been absolutely no conflict between the programs. When the update for Dragon NaturallySpeaking with Vista comes about, I suspect there will be no conflict between
Windows® Speech Recognition and Dragon. These are two completely independent programs and use completely different speech engines and user interfaces.

Martin

-------------------------

Martin Markoe - BANNED USER: This user has been banned from these forums

mmarkoe --

I have had the same experience as you. Dragon and Microsoft Office 2003 usually worked harmoniously on both my home and office computers. Sometimes I found the dictation in Microsoft Word to be initially a bit more accurate, particularly when using words that were part of my "work life". Overall, however, I found dictating into DragonPad to be much faster. It was much easier to correct mistakes. Selecting all that I had dictated (typically half a page or four to five paragraphs) then copying and pasting the material into my document became a fairly rapid routine.

But everyone has not only a unique speaking voice but also their own ideas of how programs should and should not work.

 

-------------------------

Cheerily Yours,

Tim Beckwith