Having big problems with AO on DNS Premium 11.5.  Have been using only a short while.  Have run AO twice, both with same outcome: after it ran the first time, the accuracy dropped to 20% from 90%. I fiddled around with it and suspected that problem might be in the Language Model Optimizer so i disabled it.  Somehow or another, I got it back to working OK.  Second time, just a few days ago, it ran in middle of night but when I got back to it in the morning, it seemed OK for about a minute, then accuracy plummeted to 20% to zero - that's right it wouldn't work at all.  Then after several minutes (I left it alone), it started doing weird things like pulling up a blank email.  I've been fiddling with it a few days - reinstalled DNS and NB2011, Rev B.  Also tried MS Restore, all to no avail.  I've had it; I've got issues and I know for the time it worked - and worked well- that the software was performing as I dreamed it might someday.  Since I have seen it work correctly, I know it's up to the task and i feel like the problem is something very simple but I haven't been able to figure out what.  Could it be I working (or not) in the Roaming Model and if so how would I fix that?  Also, I haven't figured out how to make sure the feature that sends old language files to archive is working - from what I've read that could be a big factor.

Logic tells me that since it worked fabulously from the get go, then started the psychotic behavior after I ran AO the first time, should I disable AO - it being "not broke?"  Sounds like a rat in the speech files.  If I ever get this thing working again, I will then have an opportunity to set it up to work right going forward and avoid making the same mistake, whatever it is, again. Is there something specific (settings, etc) you can recommend I do to set me on the right course?

The other option I've considered but haven't tried is to delete all the speech files and start over.  That seems drastic and I didn't want to try it w/o knowing the consequences.  Back through the versions, none of which have worked satisfactorily until now, I've dictated a fair amount of material into the system.  Is it possible that over the years during which I have used it very sporadically, I've built up a critical mass of core speech files that have now made it possible for the machine to work great; and if i delete them all now in an effort to start with a clean slate, am I destroying the very thing that now allows it to work well?

Any solutions you can recommend will be most appreciated.

Best,

 

And now, how do i send this thing?  I'll figure it out.

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

What does the "w/Prozac" mean???

 

R 

Thanks for pointing that out, R.  Have edited the query.


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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

From your description, the problem may not be related to running the Optimizer. More likely, there is a problem with your sound system. So here is what I would do:

1. Run the Microphone Check (aka Audio Setup Wizard). This takes less than a minute. On the final screen of the Wizard, click the "Play" button, and listen carefully to the sound of your voice. It should be crystal clear, with no snaps, crackles or pops. Rice Krispies jokes aside, you should hear no noticeable distortions, hums, buzzes, environmental sounds, or breathing sounds. If you hear breathy sounds, your microphone may be probably improperly positioned. If you hear anything other than crisp, clear speech, there is a problem with your sound system. 

2. Create a new user totally from scratch. This takes five to ten minutes. Don't worry about the speech data you have collected. If you don't get excellent accuracy immediately, something is wrong. If this resolves the problem, return to your original profile, export your custom words and commands (if you have any), and import them into your new profile.

Alan, Thanks so much for your prompt response.  I'm off to class but I'll try it tonight and report back.


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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

You can rule out Roaming because that feature is only available in DNS Pro.

Many end-users have been able to appreciate higher accuracy after running the Acoustic Optimizer but some end-users have reported similar unfortunate results. In our personal tests, it didn't make any difference to our speech files but it's obviously not working for you. In your case, you might consider disabling this feature and creating a new user. Your current profile sounds like it may not be corrupted.

Most end-users are shocked at how good a brand-new user profile works after religiously saving their user files for years and we suspect you will have a similar experience. However, it isn't necessary to throw away your previous vocabulary or commands. You cannot save your pronunciations (another DNS Pro exclusive amenity) but they are probably damaged anyway. Additionally, individual word pronunciation is overrated . Here's what we recommend:

1. Click on the DragonBar Vocabulary menu
2. Select Export Custom Word and Phrase List
3. Export your vocabulary to the desktop so that you can find it later
4. Create a new user profile
5. Reverse the process of steps #1 & 2 by importing your vocabulary via the Import Custom Word and Phrase List
6. If you have commands that you wish to preserve, close your current NaturallySpeaking user profile and copy your old~ C:\ProgramData\Nuance\NaturallySpeaking11\Users\(your username)\current\MyCmds.dat file from your old user to the new user.

HOWEVER, you may not have to go through the previous procedure because NaturallySpeaking automatically creates a backup copy of your user profile every 5 saves (default setting). You might simply try switching to your backup user profile. You will find 2 step by steps on this process in the NaturallySpeaking Quick Tips (at the top of this page) when you look up “switching” as in switching to your back end-user profile.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

Dear Alan and Lunis,

Thanks for your suggestions.  I've been playing with them all evening.  First, I tried Alan's suggestion of re-running the Mic Check, then clicking on the "play" button.  It all checked out; got clear, crisp speech.  So, from your note I assume we can eliminate a defective sound system.

Then I tried creating  a new profile from scratch. Afterwards, I have the same result, i.e., when I speak, it gives me very few words, all wrong, and sometimes weird things asking me if i want to to put the last phrase (the one I dictated, correctly) in a special vocabulary.  Another potential clue for you pundits is that every time I try to dictate something there's a small "error message" that says: "Please say that again".  Of course I do, but then get the same message. 

I'm about to lose faith in this thing again.  It was going to be my ticket for finishing a long and important school project that's due on Tuesday. It wouldn't be so frustrating if I hadn't had a taste of how helpful it can be whenever it does right.  If you can think of anything else I might try, please let me know. Meanwhile, I'll get back to my paper and hope I can finish.

Best,

Homer

PS.  FYI, I also tried resetting my mic (GN9330e USB) w/o result.  Makes me wonder if I reset it correctly.  How would I know?  FYI, the blue light on the console and headset continue to flash but from the user guide I gather it only means that the USB is on functioning correctly.

 

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

Lunis,

Although I haven't seen any evidence that my issues are mic-related, I defer to your expertise.  If you have some suspicions that the 9330e USB could be having some impact on the problem, I'd be more than willing to drop my "order" for the 6-in-1 and go with the 3-in-1.

Best,

Homer

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

Are you perhaps using a notebook computer? If so, can you verify a 2nd sound system through the NaturallySpeaking Audio Check? The reason why we're asking is because we wonder if you might be using your integrated soundcard and integrated microphone. This would give you a reasonably high quality playback and extremely poor accuracy although you might occasionally experience reasonable accuracy, such as when you move in close to your keyboard. We are just spit balling but frequent displays of the “please say that again” message translate into a substandard sound system improperly sound setup set up.

PS: We suspect that we will sell out of SpeechWare 3-in-1 TableMikes in about a week and will not be able to get in another ½ shipment until the end of March.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

Lunis,

I did as you suggested and from the info that appears (see attachment) it looks like I'm correctly plugged into the correct system.  If that's not a good assumption, please tell me what i can do to be sure. 

If it is a defective sound system, is there any other way I could confirm that.  For what it's worth, I notice when I'm dictating into the mic, that the indicator is active showing that the mic is transmitting correctly into the software. Wouldn't that suggest that the sound system is doing its job?

In any case, I've decided to order the 3-in-1 before your supply dries up. Might this device use its own sound system thus eliminating any more issues regarding the dictation problem?  Like I said, I'm ordering it anyway - I read over the marketing and technical information brochure with respect to the 3 and as updated, it seems to cover all the concerns I had, especially with the extended range.  I probably wouldn't use the pedal anyway.  Let me know how to order it please so that it includes the pricing arrangement we discussed.

Thanks again for your spitballs and other help.

Best,

Homer

 

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

1. Do you use NaturallySpeaking in a noisy environment? That could mean that you are listening to music, your neighbours are listening to music, there are people having muffled conversations nearby, or that there are appliances like refrigerators, dishwashers, fans, etc. that are running?
2. Have you tried using a different USB port for your microphone? Not all USB ports are equal.
3. Are you using a USB hub? How many devices share it with your microphone? (The correct answer is zero!)
4. Does NaturallySpeaking respond in any way when your microphone is unplugged or muted? 

Dear Alan,

I've been using a USB port on the back of my laptop (not the docking station) which understood was the best deal. In checking out your comments, I decided to try the other two USB ports on the side. I tried it immediately with no results. Then, I left my desk for about 10 min. and after returning I noticed that the software was looking a bit more perky, so I put on the headset to give it another try and voilà, it was back. It still does some funny things like, each time I say "Delete that," it activates a new tab. I'm not worried about that and am pretty sure I can fix it. The next step comes after leaving it on for a while, in checking to see that it's still working properly. Subject to that caveat, -oops, I just hit a snag: some words are being constantly missed by DNS and are stubborn about being trained.  "Oops," is a good example; it doesn't seem to mind how I train it, it just doesn't want to set it up in vocabulary. Other examples include commands, such as "Spell That," "Delete That,"and "Correct That." they all react in the same way by opening a strange and purely unrelated window. Is it possible that when I set up this new profile, it wiped out some number of spellings and commands? If so, is there any way to reverse that or should I just assume that I'm having to rebuild some lost words and commands and proceed accordingly?

Lunis, just to confirm, I'm still interested in buying the 3-in-1 now so let me hear from you on that, preferably before your supply runs out. And thanks so much to the both of you for your help in walking me through these various solutions. If this snag becomes a dam, I'll be back in touch with you.

Best,

Homer

 

 

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

We can confirm that you're using the correct sound system and even sounds like it's functioning correctly. We just can't explain your loss of accuracy.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

Alan and Lunis,

Just knew you want to know before turning in for the night that my DNS is right back where I started when I wrote my first post. 

After I finished that last email to you, still gloating!, I turned the mic off and put the computer in sleep mode.  When I came back 90 minutes later, it had regressed to the point where it was just after this nightmare started.  So, essentially it is non-functional.

If any more ideas strike you, let me hear.  Thanks again to you both.

Best,

Homer

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

The worst thing you can do with Dragon is allow your computer to go into sleep mode. Your experience is exactly what happens when you bring it out of sleep mode. If you need to put your system to sleep, use Hibernation, not Sleep mode.

Chuck Runquist
Technical Project Manager
VoiceTeach LLC
Home of VoicePower®: We don't make Dragon NaturallySpeaking, We make it better!

"Life's Rule #1: Once you pull the pin, Mr. Grenade is no longer your friend."  (Variant of Murphy's Law  -  Edward A. Murphy, Jr)

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

I'm responding to Chucker's last note about the sleep mode issue.  Thanks, it's good to have that info b/c I generally use the sleep command a good bit. Anyway, I followed your advice and re-booted the system.  For a few short minutes, it worked fairly well, then checked out completely which is where I left it just after my last post to you. A few minutes ago I decided to give it one more try before reporting back to you. Fortunately, it seems to be working again though not nearly as well it did before we had all these problems with the acoustic optimizer which prompted my first post. After using it only 10 minutes, not having used it for three or four days, it seems to be working again though not nearly as well it did before we had all these problems with the acoustic optimizer which prompted my first post. There seems to be a pattern to the mistakes it makes. There are probably more, but there are a few I particularly noticed and some seem to have a commonality (no particular order):

• doesn't recognize Select command-it simply doesn't respond. So I manually select a word and say Correct That or Spell That - no response.  However, Delete That works.  
• doesn't recognize "Bullet That" or anything else with "bullet." - Bold that instead.
• "... That" commands.  Instead, it goes to some other place.  Earlier this afternoon, with some command (bold that, as i recall) it came up with with a fresh email ready to go with the word "bold" in the subject window.
• It no longer caps the 1st word in a sentence; nor does it skip a space after punctuation at the end of a sentence and sometimes in between words.

I'm far from any sort of expert on these matters (tho before it's all over, I may have the knowledge level of a PhD) but it sounds like i did a major "clusterf _ _ k"  on my profile back when we were trying to get it to act right, resulting in more than a few corrupted files.  Here's my question, I'll be ordering a new 3-in-1 from Lunis this afternoon and when it comes it looks like the perfect time to start afresh with a new profile (speech files and all) including the backup speech files? What can I expect as to performance from the revisions and the new 3/1 after this maneuver? Make sense?  Shall i delete all the other profiles, as well? How about the backup speech files? Also, I have disabled the acoustic and language optimization functions on advice from Lunis - is that OK from the get go? I'm trying to develop some reliability with this puppy; what can I expect as to performance from the system with the revisions and new 3/1 after this maneuver? I believe its success will depend a good bit on laying a solid foundation, and hence the reason for my anal approach. But I've had a taste now and remain hopeful we can work out a way to peacefully coexist.

Thanks for your indulgence.

Best wishes, 

 Homer M

 

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

We disabled the Acoustic and Language Model Optimizer years before it existed . No, that won't cause an issue.

PS: Only 5 more 3-in-1's left in stock and then we are out until the 2nd week of April.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

I thought that Acc Optimiser was kind of irrelevant due to the silent adapt feature????

R 

"silent adapt feature."

Don't know what this is. 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

Rag,

Without going into a lot of technical detail, the SilentAdapt feature is like running general training on-the-fly. The Acoustic and Language Model Optimizer is a comprehensive adaptation of your user profile based on corrections, training, incorporating the dra files in the voice_container\drafiles folder, and the Language Model adaptation. The two are not the same (i.e., apples and oranges) even though SilentAdapt does perform basically similar to performing a general training. Also, the SilentAdapt feature uses a Nuance patented algorithm called the PelAudio Acoustic Scale Score, which assigns and updates confidence levels for words that you use most frequently. Have you ever noticed, albeit fairly rare, that you ran into a situation where a word that Dragon previously recognized accurately ends up being misrecognized. The results of the PelAudio Acoustic Scale Score over time tend to teach Dragon to ignore words that you don't use on an everyday basis, or have added to your vocabulary but not used subsequently. Some users may have noticed this and some may not depending upon how you manage your Active Vocabulary.

Regardless, why would Nuance continue to provide both SilentAdapt and the Acoustic and Language Model Optimizer if the latter were "irrelevant"?

earcoach,

Hopefully, this should answer your question as well.

Chuck Runquist
Technical Project Manager
VoiceTeach LLC
Home of VoicePower®: We don't make Dragon NaturallySpeaking, We make it better!

When you have eliminated the impossible, whatever remains, however improbable, must be the truth. - Sir Arthur Conan Doyle

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

Sorry guys but i'm confused.  First, are you saying that the "silent adapt feature" is an alternative, perhaps more benign?, than the optimizers, so that if you turn off the Optimizers, the Silent automatically kicks in and performs some role as a minimal Optimizer?  If not how do you get the the Silent to come on? 

Secondly, I'm a bit puzzled about Rag's question about the Optimizer being irrelevant. I don't believe I ever made that statement - I don't know enough about it to form any opinions. What prompted him to ask that question?  I've looked back over the discussion and can't find a reference that would suggest why he asked it?  I'm posing this question not to be argumentative but so that when I set up this new profile (By the way, please don't forget to answer those queries from last night.  I have a feeling you may think that since i purchased the 3/1, my problems are all over.  Actually, the problems are just beginning if I don't set it up right which is the reason I've peppered you with all the questions) I will have done it right.  If i don't, then I'll be right back where I started on these issues and Lord only knows the frustration it causes me and the probable irritation at your end over this fool (me!!) who can't get it straight and keeps barraging us with all these questions.  Actually, I'm trying to do us all a favor.

Thanks for your continuing patience,

Earcoach

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

After I run the acoustic Optimizer my accuracy is phenomenal just a few days later however, without a lot of dictation my accuracy seems to diminish greatly. It's a little annoying to have to remember to run the acoustic Optimizer every day or two.

Theoretically, how often should I have to run the acoustic Optimizer? I was thinking I had read on here somewhere that it was supposed to be once a week.

Wheels,

If your accuracy improves significantly after running the Acoustic and Language Model Optimizer, it shouldn't diminish even if you never run the Acoustic and Language Model Optimizer again. Something else is working behind the scenes to cause that to happen.

First, what you should do is to check the number and size of the dra files stored in the drafiles folder (i.e., current\voice_container\drafiles). If the number of dra files is significant and/or if the size of some of these is rather large, then it's time to run the Acoustic and Language Model Optimizer. However, whether or not you do after running it the first time shouldn't have any impact on overall accuracy.

Second, if your accuracy is degrading over time after running the Acoustic and Language Model Optimizer, then you should do the following in the following order:

1. Rerun the Audio Setup Wizard (Check Microphone).

2. Close and relaunch your user profile making sure that you save your profile in the process.

3. Making sure that you save your user profile, close Dragon completely and restart it.

If performing the above in the order specified doesn't rectify the accuracy situation, then running the Acoustic and Language Model Optimizer once or twice a week is not likely to improve your accuracy. Running the Acoustic and Language Model Optimizer is like taking out the trash. It gets rid of redundancies, adapts your user profile acoustically and in terms of corrections that you have made, updates the Language Model, and does a certain amount of consolidation by virtue of throwing out the trash so to speak. However, it may not make your user profile smaller, just more efficient and this is what improves overall accuracy.

Regardless, if your accuracy is degrading, then it's time to look elsewhere for the cause.

Chuck Runquist
Technical Project Manager
VoiceTeach LLC
Home of VoicePower®: We don't make Dragon NaturallySpeaking, We make it better!

"What you are aware of you are in control of; what you are not aware of is in control of you." - Anthony de Mello

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

Chuck,
Thank you for the advice. I will report back in a couple of days, when my unfortunate usual drop in accuracy occurs. I actually had a pretty decent idea of what the acoustic Optimizer did. That's why it had seemed weird to me that it fixes my issues. Or at least seems to fix my issues.

 

My accuracy degradation has not yet happened. But I'm interested in finding out why it happens since it does happen fairly often.

I remember reading on the form that it is not ideal to put the computer to sleep. I do this fairly often but I usually make sure to turn the microphone off. When I do this with the microphone asleep or running I run into all kinds of problems. Could this practice because in my issues.

Also I do not routinely save the changes to my user profile. Once I get a profile working I typically find that saving changes only causes me to lose an accuracy. I'm thinking this is because when I make changes to my user profile, the misrecognitions were actually poor dictation on my part. Could this also be causing my routine drops in accuracy?

As you can imagine is very annoying to have had good profile occasionally but to routinely lose accuracy. Especially because I need to rely on Dragon for importing large amounts of text.

Whenever you notice a drop in accuracy, the first thing you should do is to take a deep breath, select the phrase or sentence and redictate it again, with just a little more caution being paid to your enunciation. I would bet that under most circumstances, you will get it right the next time.

It certainly has nothing to do with running the optimizers ever so often, accuracy rising soon afterwards, just to drop again a few days later.

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

I will definitely take that advice to heart. I am usually very careful about my enunciation. Sometimes I swear it doesn't make a difference but as they say the devil is in the details. Possibly my enunciation isn't just a little bit off and I do not realize it.

 

All of this being said. Can anyone recommend things I can do to work on my enunciation? Exercises I can do or things I could practice.

I recognize a drop in accuracy when I'm fatigued, especially at the end of a long day at work, interacting with many different people.

I see a greater drop still when for whatever reason my mouth gets dry.  Keep water handy - clears it right up.

Not much of an offering here but i hope it helps.

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

Don't get me wrong, but the best of practicing this is to let Dragon be your teacher. It really doesn't take anything peculiar to take care of other than just speaking clearly and consistently. The most difficult situation to dictate in, however, is when you feel you don't have the time for it. Trust me, I know what I am talking about, using Dragon in my job every day.

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

thanks for the tip. That definitely makes sense, I will give it a try.

you seem very knowledgeable and by no means was I doubting what you were telling me.
Thank you to you both for the great advice.

It wasn't my understanding that you were doubting me, what I meant to say was that, if you are using Dragon under working conditions, your accuracy will alternate to any kind of degree over the course of a day quite naturally, just like the shape you are in will also. Unless you speak like a robot.

Just one more note, are you aware that the human voice has a lot in common with the human face, as well as peoples' gestures? Please remember that Dragon can read your mind by listening to you, and while you can buy all the microphones in the world (sorry, vendors), when talking to him, he wants to see you relaxed, smiling, and being happy.

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

Thanks for humoring me and for the advice. I have noticed that my disposition affects the way my dictation works. I suppose even a slight change in the tone of voice can make a large difference. I guess I will just try to remain relaxed and happy while I am dictating.
Thanks

That's what most of it boils down to. The impact of this has even increased in version 11, and that's probably one of the reasons why some people have a hard time moving from 10 to 11.

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

It's interesting that you mention it. When version 11 first came out everyone was talking about how much more accurate it was. I didn't notice any difference whatsoever. However, I can't remember if I was using a crappy headset with version 10 and moved to a better headset with version 11.

Thankfully I also went from a home to a premium version with version 11.

It's interesting that you mentioned the differences in version 10 and 11.

When version 11 the first was released everyone was talking about how much more accurate it was. Unfortunately, I don't think I noticed any accuracy difference.

 

I am not completely certain if I was using a crappy headset with version 10 and shortly after adopting version 11 I switched to the boom O.

I also think one of my major problems is forcing myself to dictate the way I know I should in order to receive optimal results. I suppose dictation is just as much mental as anything else.

 

In my opinion, it is actually, but you need to treat it a little differently from what you were used to up until version 10. I can't put my finger on it exactly, but the downside to this, it is also less forgiving at the same time.

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

I basically find the same difference between DNS 10-10.1 and DNS 11-11.5.

DNS 10-10.1 is more forgiving of slurred speech, running your words together, etc.

DNS 11-11.5 is more accurate, but it is more sensitive to changes in dictation style. It's less forgiving in situations where you slur your speech or run words together. It seems to require more precise enunciation and responds better to using a more smooth and natural dictation style. It's like the primary script for talking to a computer. In DNS 11-11.5 you really do have to more closely mimic the style used by newscasters.

Improvements in accuracy seem to revolve around more accurate responses to things like saying "numeral one" and words and phrases that DNS 10-10.1 had trouble with. For example, when you use the "Use alternate written form 1, DNS 11-11.5 seems to pick it up right away vs. DNS 10-10.1 which often does not. Custom words added to the vocabulary seem to respond better in terms of accuracy on first use. The Language Model seems more accurately assess context, thus producing more accurate results, particularly with a longer phrases.

On the other hand, DNS 11-11.5 is more sensitive to lower quality microphones and performs better with the higher quality microphones in terms of overall accuracy.

I could go on but in my opinion Rüdiger seems to have generally nailed it. It is difficult to put your finger on exactly when and where DNS 11-11.5 is more accurate and where it is not. However, overall if you have a smooth, natural, dictation style in which you clearly separate your words and clearly enunciate, and you're using a good high quality, noise canceling microphone, that it generally tends to be much more accurate than DNS 10-10.1. However, it is difficult to pin it down precisely and specifically.

Just my general observations.

Chuck Runquist
Technical Project Manager
VoiceTeach LLC
Home of VoicePower®: We don't make Dragon NaturallySpeaking, We make it better!

"In theory, there is no difference between theory and practice. But in practice, there is." - Yogi Berra

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

Chuck and everyone,

just today, I was lucky to make quite an interesting observation as far as the differences in overall functionality and specifically in accuracy between those two versions. In the office, I had to go back to a number of dictations, which I had recorded in the field in January 2010, using the Olympus DS-5000. Back then, the recordings had been downloaded to our servers, and initially I had started transcribing a significant part of it into DNS 10.1 at that time, thus saving the DRA files (text along with audio) next to the original sound files. It turned out that this particular project I was working on had been delayed for when we "get around to finishing it", so we put it on the shelf, and kept the files on the servers, with no further editing done to the files, but just leaving them as is after transcribing.

So you might think that from then on, we kept sitting around in our offices with our feet on the desks, but that's just a part of the truth.

Just the other day, we were reminded that something is still missing (not really unusual, there are always certain bits that we are more or less a couple of years behind our schedule with, but that's just because we get such loads of new orders every day that need some attention right now), so to make a long story short, I dug out all those old drafts and DRA's just to find out what kind of silly mistakes there were all over the transcriptions done with 10.1. Things I had long since forgotten that they may occur, but were quite the fashion during those days.

Equally interesting maybe, correcting them was just a piece of cake, by selecting and dictating over - using 11.5 it worked at first attempt in almost all instances.

So, you may look at it whatever you like, but I am definitely not looking back.

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

Although accuracy is an important component when comparing microphones, sound cards and speech recognition applications, what really matters in the end is productivity. Ver. 10 and Ver. 11 each have their own unique problems such as the not ready for prime time Ver. 11 Nothing But Words algorithm, which would be awesome if it were finished. For us, Ver. 11 introduced a lot of new time-saving features including the following example:

1. Dictate the following sentence in Microsoft Word, DragonPad, WordPad, WordPerfect, Internet Explorer or NotePad: This is what I said and this is what I meant.

2. Use your mouse to select the letter n in the word and as follows: This is what I said and this is what I meant.

3. Dictate period and prepare to be impressed as NaturallySpeaking removes the entire word, as opposed to dictating over a single letter, inserts a period to the immediate left of the previous wording and capitalizes the 1st letter of the next word. This is just 1 of numerous new time-saving features. You'll find a few more in our now somewhat dated NaturallySpeaking 11.5 Review.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

now that I think about it. Agree. I think a large part of my problem could be that I slur and do not always enunciate clearly.
I also tend to speak in a bit of a flat fashion. Which I guess I was able to get away with in version 10. But, may be causing me problems in version 11.

Lunis,

Presumably you've seen the overnight sales since you got in.  Sales order #: IVA4667011675.

Have an email sez it shipped last night.  Please check and confirm.

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

Earcoach,

Your order will theoretically be picked up in a few minutes by FedEx Express Air but considering that we are in the middle of several twisters, your package may go airborne a bit prematurely

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

 Ya'll,

Last time I posted I had just ordered the 3-in-1 desktop from Lunis.  It's been here a while and I'm just getting around to installing it. As you may recall from our exchange, I had been having serious, continuing problems getting DNS 11.5 w/ KB2011 to recognize and correctly transcribe my voice. Lunis suggested it might be a corrupted file problem and recommended i set up another user with a clean slate which i did.  It worked OK for less than a day then went right back to the same problems I'd been having (for more details see the msg traffic). I ordered a 3-in-1, not to solve the problem, but as a high quality enhancement. It seems to me this is the perfect time to start afresh with re-installation of DNS, KB2011 according to instructions in the KB Handbook, and the new 3-in-1. My plan after loading the basic software is to install the new mic.  This is the point where I inserted the message to Chucker quoted below:

 

These questions went unanswered pending my receipt of the mic and set up time.  Now that we're here, I'd like to re-pose these questions so I'll know exactly how to handle the issues during the installation. Essentially, where I'd like to go, subject to your advice, is to install the whole system as though I were a 1st time user and just opened all the boxes (in fact, the relevant software has now been uninstalled and I'm ready to get started [is there any need to try and find some software right quick that will clean up all the residual files left by DNS/NB]?).  Now, please answer the questions from the quote.  I'll try it and let you hear back.  

 

Two more small questions:

• I have the 11.5 update for DNS (separate sw I believe).  Should I install the ND2011 after the 11.5?  I assume so, but please confirm.
• When I'm setting up the mic and speaking into it for purposes of sound, etc, where do i hold it relative to my mouth?

Thanks gents,

 earcoach

 

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

DNS11 Remover is designed to cleanly remove every trace of NaturallySpeaking from your computer but it will also destroy your user profile and in your case, it shouldn't be necessary anyway because you're about to reinstall NaturallySpeaking. In other words you shouldn't need to run this utility although it might be a good idea to download it for possible future use.

No need to worry about what order everything should be installed in or how to install your upgrade as a full copy because we just sent you the January 2012 edition of our new Installation/Training Guide which includes this information along with our new DNS Beginners, Intermediate and Advanced Training. Note that this guide is available to any of our DNS clients on request.

Microphone positioning depends on your vocal strength but we typically recommend 6 inches when running the Microphone Check. Once you have completed the Audio Check, feel free to extend the distance up to 24 inches.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

Ya'll,

I’ve completed the reinstallation of DNS and KB 2011; everything seems to be working fine. It’s not quite as accurate as it was but I guess that’s the price you pay when you cancel out all the speech files and start from scratch. I chose the option not to use the accuracy/Optimizer feature - from our previous conversations I gather that was okay although I’m not sure how the software can ever be any more accurate than it is now (which ain’t too bad) if you’re not saving speech files so that the machine can take advantage of its mistakes and corrections.

Thanks,

 

earcoach

 

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

Guys: I'm back in the soup!  MY WORSE FEARS HAVE COME TRUE.

 

DNS quickly reached a point where it is not transcribing at all. Started getting a continuous audible alarm with message saying "No speech available".  At another point, the message was "cannot connect to DDE Server App." At still another, it was that i didn't have enough memory available (that seems crazy with the 12GB memory component I have). I'm working on a school project with deadline looming so please help before you shut down for the weekend.

As always, many thanks.

Earcoach

 

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

EarCoach, loyal KB customer

DNS Premium 11.5; Win 7 Prof, 64 bit, SP 1;

RAM:16 GB; Intel Core i7 CPU; 1.73 GHz; 

Mic: GN9330e USB 

Quote:

Please remember that Dragon can read your mind by listening to you, and while you can buy all the microphones in the world (sorry, vendors), when talking to him, he wants to see you relaxed, smiling, and being happy.

Good point. We are now considering selling alcoholic beverages



Most, if not all, of this problem is associated with Windows; try rebooting.

Last note: Unfortunately, there are a surprisingly high number of reasons why your accuracy can drop and here's 1 that you may not have thought of:

For test purposes, we occasionally import a medical vocabulary which can range from 9000 - 15,000 words. We usually remember not to save our user profile so that the vocabulary doesn't become permanent but we blundered about a year ago and made 1 of our medical vocabularies a permanent part of our personal vocabulary. We probably experienced a temporary drop in accuracy when we 1st did this but over time, our accuracy rose to 99%. We've been putting it off but last weekend, we painstakingly removed the medical vocabulary by selecting all of the added words and un-highlighting 3000 of our personal words and phrases. This procedure cost us some of our personal vocabulary but for the most part it was a success. HOWEVER, our accuracy dropped by about 2 - 3 percentage points and although it is steadily improving, it isn't quite back to what it was a week ago. Although this has nothing to do with your current issue, we have heard from a number of end-users who thought they could improve their accuracy by deleting 50,000 - 100,000 seemingly unnecessary words and wound up killing their accuracy. We just thought we would share this information for those who find it interesting.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

I didn't know that. Why the difference?

-------------------------
Art Silen

artsilen@sbcglobal.net; art.silen.mediator@gmail.com  

"Question assumptions"

DSN 11.5, running on Win7;Intel Core i7 CPU 970 @3.20 GHz; System Memory 12 GB; Graphic AMD Radcon HD6570; System Board 2A861.04E01; System Bios 6.15

I am posting on this thread as I seem to have just encountered similar problem and am at a loss as to what to do. Previously DNS working just fine. Ran the AO.  My accuracy dropped to 0%.  Rebooted system.  Re-ran microphone setup.  Microphone is crystal clear (Samson Airline 77). Ran "Repair" of Dragon from Control Panel.  No good.  When I dictate a simple phrase "our training program is very broad and comprehensive", I get  "our you are you are".  When i highlight this phrase and click "Playback", it sounds very clear and just what I dictated - just incorrectly transcribed. This occurs whether I am dictating in Word or Outlook. Any suggestions?

This is why we typically warn against blindly running the Acoustic Optimizer. For some people, the effects can be disastrous, although this is fairly rare. It sounds like you have effectively corrupted your user profile. In the future, you might consider making a backup copy of your user profile before running the Acoustic Optimizer or even just not running it.

Fortunately, there is a good chance that you can switch to your backup user-file if you haven't saved your current profile too many times (which overwrites the backup user profile every 5th time). You'll find a step-by-step on switching to your back end-user profile when you search for “switching” in the NaturallySpeaking Quick Tips (at the top of this page) as in switching to back up user files.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

The backup version was slightly better but still significantly impaired. I ended up creating a new profile and we seem to be back in business. The take-home message seems to be that I should not run AO manually. If so, lesson learned. Thank you, Sean

smackey,

First, I'm in no way criticizing Lunis with regard to his feelings about running the Acoustic and Language Model Optimizer. Some like it, some don't. Some think it performs its functions properly, and some don't. Generally speaking, however, in of itself the Acoustic and Language Model Optimizer does not damage a user profile unless the user profile is already damaged and/or corrupted. Drawing the conclusion that because your user profile doesn't function properly after running the Acoustic and Language Model Optimizer is generally not valid. Before running the Acoustic and Language Model Optimizer you should always export your user profile using the Manage User Profiles… (DNS 11-11.5) or Manage Users… (DNS 10-10.1) | Advanced button | Export. Then, if something goes wrong with the Acoustic and Language Model Optimizer and your profile doesn't function properly, you can always Import your exported user profile. However, the fact that you had a problem with your user profile after running the Acoustic and Language Model Optimizer is a good indication of the fact that there was something wrong with your user profile to begin with. Not manually running the Acoustic and Language Model Optimizer is throwing the baby out with the bathwater there is nothing inherently wrong with the Acoustic and Language Model Optimizer such that it would do what it did in your case unless there was a problem with your user profile to begin with.

Second, you indicated in your post that even your backup user was significantly impaired. The lightbulb should have gone on at that point because if your backup user is significantly impaired, that's a good indicator of the fact that your user profile was damage to begin. The reason that I point to that is that the Acoustic and Language Model Optimizer does nothing with regard to your backup user. It only acts on your active user profile, it does not optimize or act on your backup user profile at all. So, if your backup user profile had a problem, then your user profile was probably corrupted in general to begin with.

My recommendation is not to avoid running the Acoustic and Language Model Optimizer manually but to make sure that you got and exported user profile before you run it. And, if the Acoustic and Language Model Optimizer results in the problems that you were experiencing, then the likelihood is that your user profile is corrupted anyway and you should create a new one at that point as you did.

Chuck Runquist
Technical Project Manager
VoiceTeach LLC
Home of VoicePower®: We don't make Dragon NaturallySpeaking, We make it better!

Logic 101: Post Hoc fallacy - The logical fallacy of believing that temporal succession implies a causal relation.

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

I appreciate that an inference could be made that, because my backup was also bad, that my user profile was bad before I ran the A0. I had considered that. However I was under the impression that the system automatically backs up your user profile every five times or so. During the issues I was having with my profile, I may have opened and closed DNS more than five times. Therefore, I wasn't convinced that the backup reflected a state before running AO. I appreciate that I may be misinterpreting the automatic backup process and that my backup may in fact have been as you said. 

In any event, it sounds as if there is some degree of controversy with running AO.  What I am taking away from this is that if one does run AO, that the profile should be first exported. Thank you, Sean

That's just about the take-home message. Just adding to it, don't take anything for granted. No one really knows what the optimizer really does and how it works - at least no one around here, despite all the gossip about it -, so you had better take care and precautions before running it.

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

Please bare with me a moment, as I grouse. Several years ago, I started taking helicopter flight lessons. I had a little time in fixed wing, and thought that would help with rotary flight. I never will forget what my helicopter instructor said, "to compare flying a helicopter to flying an airplane, is like comparing riding a unicycle to riding a bicycle. Son they just ain't the same thing".

I am beginning to think that comparing voice dictation to typing is about the same thing, except you do not get killed if you mess up. When I was in high school in 1964 I had the opportunity to take typing. At that time, only girls and guys who acted a lot like girls, took typing. Sometimes I wish that I had acted a lot like a girl and taken typing.

I too started several weeks ago with very good accuracy and amazement when I used DNS 11.5. Then I decided to improve things. So I ran the Acoustic Optimizer, AO, and just like several others (and I wonder how many more that have not posted on the forum yet), I found that my accuracy has plummeted. So, I bought a new Samson Airfone 77. $300 later, I do not see any improvement over my $75 Andrea ANC 700, (and it is certainly not as comfortable).

I have run into the problem that some others have mentioned, such as dictating a word and it goes to a command. As an example, when I started to dictate the first three words of this post, "Several years ago", a new e-mail was opened, taking the dictation as a command. Now where did that come from? As a matter of fact, I am dictating this in DragonPad, and with fairly good accuracy. When I tried to do the same thing directly on the forum, it was awful.

I think that I have been struggling with Dragon NaturallySpeaking since version - 0.0. Every new version was supposed to be better than the previous, and I acknowledge that some if not most were. But the frustration that goes with the new versions and even degradation of the existing version at no fault of mine, is just that, frustrating.

I feel like I need to have my diaper changed, fed and put to bed. "Wah, wah". There, now feel better!

Other than taking typing lessons, any suggestions?

-------------------------
1NcaNon

If you are using Internet Explorer, are you remembering to launch it after NaturallySpeaking? Other browsers do not include NaturallySpeaking support and by default, if NaturallySpeaking thinks you are in a non-Full Text Control application, it will take anything you say and attempt to turn it into a command to search the Internet, your computer or send an email.

If you open your DNS options to the Commands tab, you will see a number of Search checkboxes which you can uncheck to tame the Dragon, so to speak. Ver. 12 includes even more options along with a prompt to allow it to mini-optimize your user profile on approximately every 5th save. We are still on the fence about making this recommendation so for the moment, we are bypassing the question because we did experience a little trouble with letting NaturallySpeaking optimize our profile but in all fairness, we are using a Ver. 12 prerelease. We have always been against optimizing but a lot of end-users have reported positive results.

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

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012
Click SpeechStart+ to Download a 15 Day Evaluation of SpeechStart+

This is a tangent, but I'm so jealous! I've always wanted to fly a real helicopter. I've had some fixed wing instruction but now want to get into gliders instead. I will say from flying RC helis (collective pitch) that once they get moving in fast forward flight, they behave very much like planes just with a lot of rudder input required. But, that's mostly from flying acrobatic RC airplanes/helicopters. I'm sure in real life manuevers, helis feel/behave nothing like planes.

Anways, you are awesome and I am envious!

 

Max Roth

Freesr.org 

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

ErgoArchitect Assistive Technologies - 

ShowNumbers Plus! Addon to Naturally Speaking - 

www.ergoarchitect.com 

 

It is possible to hose a profile, although there is no consensus on how to do it. Here are some theories:

1. Running the Acoustic and Language Model Optimizer.
   
2. Allowing a computer enter Sleep mode while DNS is running.
   
3. Shutting down a computer without shutting down DNS.
   
4. Living in a universe governed by the second law of thermodynamics, i.e., over time, everything tends toward disorder.
   

For me, running the Optimizer doesn't cause harm, and may in fact incrementally improve accuracy. I sometimes wonder whether the Optimizer hastens the demise of profiles that have begun to develop problems caused by, for example, letting the PC sleep while DNS is running.

 

My DNS profiles do not last forever. I don't know why they degrade, but it happens less often now that I am in the habit of avoiding sleep mode and exiting DNS before shutting down. But there is no escaping the tendency of everything in the universe to break down over time!

That's another case of entropy manifesting itself. All silliness aside, how could even expect something to further improve that's reached its climax pretty soon already, by trying to "optimize" it? The guy who first came up with the term "optimize" should be strangled in retrospect in my opinion. Trying to optimize something that's optimal the way it is can only have a detrimental effect.

Rüdiger


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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

"All silliness aside, how could even expect something to further improve that's reached its climax pretty soon already?"

Rudiger,

Viagra

MDH

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

Mark,

 

greetings to Freud then? Maybe I should have said "top-level" instead of "climax"? Please help me with this, being a non-native speaker.

 

Rüdiger

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!

 One option would be to say: "All silliness aside, how could you expect to further improve something that is top-notch already?"

-------------------------
Crohn's News Blog: www.crohns-news.net

 

Thanks for this. That would sound more like it, of course.

 

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

Well, it's past the point where we can make any changes in the code, but we can still make changes to the Easter Egg!