KnowBrainer Speech Recognition
Decrease font size
Increase font size
Topic Title: Excessively large User Profile
Topic Summary: Step by Step for Dummies how to trim a large User Profile
Created On: 04/08/2016 12:23 AM
Status: Post and Reply
Linear : Threading : Single : Branch
 Excessively large User Profile   - Lucien - 04/08/2016 12:23 AM  
 Excessively large User Profile   - Stephan Kuepper - 04/08/2016 03:11 AM  
 Excessively large User Profile   - Lunis Orcutt - 04/08/2016 01:23 PM  
 Excessively large User Profile   - Lucien - 04/10/2016 09:05 PM  
 Excessively large User Profile   - Lunis Orcutt - 04/10/2016 09:35 PM  
 Excessively large User Profile   - Stephan Kuepper - 04/11/2016 02:43 AM  
 Excessively large User Profile   - Lucien - 04/12/2016 03:04 AM  
 Excessively large User Profile   - Lunis Orcutt - 04/12/2016 12:17 PM  
 Excessively large User Profile   - David.P - 07/04/2021 08:25 AM  
 Excessively large User Profile   - Lunis Orcutt - 07/04/2021 01:35 PM  
 Excessively large User Profile   - David.P - 07/04/2021 01:58 PM  
 Excessively large User Profile   - Stephan Kuepper - 04/12/2016 05:16 AM  
 Excessively large User Profile   - Stephan Kuepper - 07/05/2021 05:49 AM  
 Excessively large User Profile   - David.P - 07/05/2021 06:04 AM  
Keyword
 04/08/2016 12:23 AM
User is offline View Users Profile Print this message

Author Icon
Lucien
New Member

Posts: 3
Joined: 04/07/2016

Hi there!

Pretty new to computers and using DNS Pro Individual 14, Core 15 processor, 4GB ram with Windows 7 Home Premium.

I have been using MS Word 2003 to dictate into instead of using Dragon Pad.

I save the User Profile with the .doc still open. I either then save the doc but usually discard it. Then I close DNS.

I have been training DNS occasionally over some months to make it more accurate. I just enjoy seeing it run so well for now.

After some short sessions of dictating and corrrecting, getting DNS to "Learn from Specific Documents", adding custom words and commands etc I ended up with a 2.4 GB User Profile.

Then I did a one - two hour dictation from a published conversation transcription between a few people with many short sentences (+ some long) with copious repetitions of the same names followed by semicolons.

I ended up with a 7GB odd User Profile.

Another session later gave me a 16GB User Profile.

Finally I could not copy the Profile to a USB drive or external HD as the computer hung (could be the computers fault, has it has done it sometimes with other applications). There could have been an error message as well but that may have been due to another issue (sorry, can't remember).

I have read a couple of the excellent posts (to users Slayer, Bradco) here about large or swollen User Profiles but couldn't follow most of it [accessing  the *dra files in Dragon Pad (none there probably because I use Word docs for transcribing into), finding the drafiles folder to delete the *drafiles and voicecontainer folder to delete the "drafiles.ini" file, locating the User Profile directory etc].

Anyway, following Chuck's advice I opted for running the Acoustic and Language Optimizer again to reduce the size of the User Profile but alas, the User Profile remained the same size.

I have always kept the "Conserve disc space... for portability") box in the Data tab unchecked for Optimizing purposes (I guess checking it now won't help reduce the 16GB at this stage).

I guess I could revert back to a previous state of the User Profile saved on another computer (most likely the 7GB version) but I hate to waste the training and the problem is likely to persist nevertheless as I continue using DNS.

I do not need to save any audio files etc for retrieval purposes at any point as far as I see myself using this program anywhere in the future.

Saving the Word Doc dictations will be all I will be doing.

Saving the training/corrections in DNS for accuracy would be helpful, though.

I hope someone would be kind and patient enough to guide me through step by step to help resolve this issue, such as exactly how to access and delete any excess files if need be, etc.

Thank you so very much.

 

 04/08/2016 03:11 AM
User is offline View Users Profile Print this message

Author Icon
Stephan Kuepper
Top-Tier Member

Posts: 2164
Joined: 10/04/2006

Hi Lucien,

I don't have time for a step-by-step-guide, all the more so since trimming a profile down to essentials is always an individual thing, but here are some general considerations.

1. Training is overrated. With out-of-the-box recognition being what it it, there is little room for improvement.

2. Optimisation is grossly overrated. Don't bother with it, the effect is minimal at best.

3. There are 2 types of DRA files: One is the dictation stored with a document. You only need those files when you want to do deferred correction, i.e. correct recognition errors in a different session from the original dictation session. Moreover, these files are stored in the same place as the documents, not in your user file.

Type two is stored in your user file, typically in subfolders of the voice_container folder. You can safely delete all these files and subfolders. Dragon needs them for optimisation, which, as I said, is grossly overrated.

4. Running the optimiser will convert the type 2 DRA files into a different format (*.nwv) and leave them in the voice_container folder, but not delete them. Delete them by hand.

5. Form these points, there follows the conclusion: Uncheck the "Store corrections in archive" and check the "Conserve disk space" buttons.

6. A profile should never reach 16 GB from just a few hours of dictation. Therefore, I suspect that, with best intentions and possibly following some half-digested forum information, you did something that caused files to accumulate in the wrong place.

My best advice at this point is: export your custom words and commands, create a new profile, import custom words and command, discard the old profile and use the new one. Loss of accuracy should be minimal. In the new profile, uncheck the "Store corrections in archive" and check the "Conserve disk space" buttons. Make a backup every now and then if you wish. Profile size should remain well under 1 GB.

Hope that helps, Stephan

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

www.egs-vertrieb.de - www.spracherkennungscloud.de

 04/08/2016 01:23 PM
User is offline View Users Profile Print this message

Author Icon
Lunis Orcutt
Top-Tier Member

Posts: 39191
Joined: 10/01/2006

                    Welcome to the World's Most Popular Speech Recognition Forum

We concur with Stephan's recommendations. It's easy to over train Dragon but even with our 8 sound sources, our nearly 3-year-old Dragon 13/14 (never trained or optimized) user profile is only 5.7 Gb. If you have a mid-continental US/Canadian accent, we no longer even recommend training Dragon. As Stephan noted, we also found very little value in extensive training and optimizing. The only thing we have to add is if you decide to prune and continue using your current user profile, rather than creating a new profile, which we also recommend, you can optionally delete your C:\ProgramData\Nuance\NaturallySpeaking14\Users\(your username)\current\backup folder. The backup folder is a carbon copy of your current folder and will re-create itself, after you save your “current” profile 5 times.



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

Change "No" to "Know" w/KnowBrainer 2020
Trial Downloads
Dragon/Sales@KnowBrainer.com 
(615) 884-4558 ex 1

 04/10/2016 09:05 PM
User is offline View Users Profile Print this message

Author Icon
Lucien
New Member

Posts: 3
Joined: 04/07/2016

Thanks Stephan and Lunis!

Perfect replies from both of you.

Somehow I could not find the folder Program Data neither any User folders on my computer in C: drive under Nuance and Dragon Naturally Speaking v14 but could see them on a previous exported User Profile folder backup on a usb stick.

Actually before I read your posts I did delete there what I thought were superflous folders/files and then imported it back to the computer to overwrite the 19(!) GB User Profile.

The program then still worked perfectly!

From what I remember yesterday it was still around 2GB in size but armed with your succinct info I will confidently go ahead and prune it back more!

Also, what I found was that the user name folder was again within another (same name user folder) when I had exported it to usb and that may have duplicated again when I reimported it to another computer and again exported/reimpoted it back a few times (if that makes any sense!).

That have been a major reason for that 19GB User Profile.

Thanks for coming to the rescue and revealing some of the practicalities such as the over-ratedness of Optimizing and Training.

Also the info on the (in my case superfluous) *.nwy files and finally now I know that I will have a lean, trim, fully trained User Profile I can go ahead and uncheck the "Store corrections in Archive" box and check the "Conserve Disc Space".

Maybe I will even just go ahead and export Custom words and Commands as suggested, make a new Profile and import them.

Whatever, problem solved and it was fun finding out the ins and outs of Dragon.

I must admit could not believe just how accurate Dragon was (as you said) initially straight out of the box , but I couldn't resist having a go at trying to make it even better.

In fact, there were a lot of Sanskrit words from various texts I was dictating from which I wanted to add to the vocabulary and commands so that may have fuelled my concerns to go further than necessary.

I had a version of DNS on another computer about 7 years ago which maybe wasn't so accurate but eventually it got so "heavy" with excessive training and saving files that the computer hung and I eventually gave up.

Again, thanks for coming to the rescue and sorry to make you reiterate it because of my not being able to find the info on your other posts .

Gratefully, Lucien




 04/10/2016 09:35 PM
User is offline View Users Profile Print this message

Author Icon
Lunis Orcutt
Top-Tier Member

Posts: 39191
Joined: 10/01/2006

You'll find 23 recommendations in the KnowBrainer Tweaks. 1 of those recommendations is to change Windows silly default settings to hide hidden files and folders in Windows Explorer/File Explorer. Your user folder address really is located at C:\ProgramData\Nuance\NaturallySpeaking14\Users\(your username) but you won't find it until you change Windows default options to show hidden files and folders.



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

Change "No" to "Know" w/KnowBrainer 2020
Trial Downloads
Dragon/Sales@KnowBrainer.com 
(615) 884-4558 ex 1



 04/11/2016 02:43 AM
User is offline View Users Profile Print this message

Author Icon
Stephan Kuepper
Top-Tier Member

Posts: 2164
Joined: 10/04/2006

Originally posted by: Lucien  Also, what I found was that the user name folder was again within another (same name user folder) when I had exported it to usb and that may have duplicated again when I reimported it to another computer and again exported/reimpoted it back a few times (if that makes any sense!). That have been a major reason for that 19GB User Profile.

 

That's exactly what I suspected. Dragon lures you to think that you should backup your profile in a folder with you name, when you should choose the higher-order folder. You're not the first to fall into that trap. Glad we could help out.

 

Stephan



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

www.egs-vertrieb.de - www.spracherkennungscloud.de

 04/12/2016 03:04 AM
User is offline View Users Profile Print this message

Author Icon
Lucien
New Member

Posts: 3
Joined: 04/07/2016

One last thing if I may ask. as a point of interest, just so that i know if unnecessary folders and files reoccur in the future:

Under the Users/(user name) folder there are still six folders left after my pruning plus a separate audioin.dat file:

1)backup2 with subfolder counters and audioin.dat file (older date from the others below)

2)chkpt with subfolders chkpt08, chkpt09 and chkpt12

3)counters with a file counts-2016-04

4)current (with other files including 3 voice_container folders from which I deleted all the dra files and *nwy files)

5)dataarchive (empty)

6)tb with subfolder counters (containing files counts-2016-01,counts-2016-02,counts-2016-03, counts -2016-04) and an audioin.dat file

7)audioin.dat file (same as the one in the tb folder)

Which of these might be redundant or unnecessary?

I just thought to back up whatever that I still have just as a precaution aside from backing up custom words and command (which I have done separately already); then creating a new User profile and importing them back in.

Thanks and forgive me for making this last query.
Lucien

 04/12/2016 12:17 PM
User is offline View Users Profile Print this message

Author Icon
Lunis Orcutt
Top-Tier Member

Posts: 39191
Joined: 10/01/2006

We concur with Stephan but 1 folder caught our eye. The tb folder should be empty because “tb” stands for temporary backup. You can safely delete the contents of the tb folder or even the tb folder itself because Dragon will create a new 1. If you have data in this folder that cannot be deleted, it likely means that Dragon cannot save your user profile.



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

Change "No" to "Know" w/KnowBrainer 2020
Trial Downloads
Dragon/Sales@KnowBrainer.com 
(615) 884-4558 ex 1

 07/04/2021 08:25 AM
User is offline View Users Profile Print this message

Author Icon
David.P
Top-Tier Member

Posts: 617
Joined: 10/05/2006

Hello @all,

 

because my user folders have grown into the half-dozen gigabyte range in size, I had a look into this issue.

 

Here's the original state of one of my user files, at about 5GB -- with only a "current" folder and not even "backup" folders included:


(The highlighted areas is what Dragon actually uses. Everything else is basically garbage or duplicated data)

 

After deleting the "tb" folder (temporary backup) and the "chkpt" folder (checkpoint, probably something similar), 1.6 GB remain:

 

After further deleting all drafiles, as well as all old subfolders in the "current" folder (there were only a few current folders and a great many old subfolders, some of them several years old), only just under 200MB remain, with the user files still fully functional:

 

That's a reduction of 96% in size.

 

A similar picture results if the contents of the profile folder are not cleaned up by file age and/or backup property, but when only exact file duplicates are identified, using an appropriate tool. In my case, the user profile contains no less than 810 exact file duplicates with a total size of 4.4 GB. Note that these are not even different file versions with different contents from backups etc. 

 

This still corresponds to 78% of unnecessarily stored, since duplicated, files.

 

Experiments like this should of course only be made if a full backup of the user data is in existence. In particular, deleting older files, especially drafiles, can of course lead to the possibility that training of the user files can no longer be done, should this be desired.



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

Sennheiser MKH Mic
Visual & Acoustic Feedback + Automatic Mic Control



 07/04/2021 01:35 PM
User is offline View Users Profile Print this message

Author Icon
Lunis Orcutt
Top-Tier Member

Posts: 39191
Joined: 10/01/2006

We have been working on this problem for several months and have found a solution that's working for us. Our end-user option changes work well enough that we updated the KnowBrainer Dragon Installation/Training Guide, which is complementary to our Dragon customers. Anyone who already has our manual only needs to click on their permanent PDF hyperlink. Our user profile, which includes the backup, is now 291 Mb and we don't believe any user profile should be more than 500 to 800 Mb unless you have multiple sound sources; which should also be avoided unless necessary. We have a separate profile for testing sound system so that we don't overload our daily user profile.

Our user profile gradually slowed down until it took about 5 to 7 seconds to drop our dictation into the target application. This efficiency decrease took about 4 hours and required closing and re-launching Dragon. It should also be noted that we were experiencing this problem on a monster i9-10980XE CPU with 18 Cores, SSD and 64 Gb of RAM. Some of our customers have reported up to a 2 minute lag. Obviously, even 4 seconds, is too long. Here's what we did…

We began by returning all the features that we originally bypassed. Because we are using 64 GB of RAM, we set our Options/Data Disk space reserved for playback to 0. You can probably get away with 0 with 16 Gb of RAM. The 2 major changes were to open the Dragonbar Options/Data tab, remove the checkmark from Store corrections in archive and add a checkmark to Conserve disk space required by user profile. On the downside, you will not be able to optimize your user profile. In order to continue using our current profile, rather than creating a new profile, we additionally deleted the %LOCALAPPDATA%\Nuance\NS15\Users\[our username]\current\2_1_Container\drafiles folder. We believe this is a Dragon bug that appears to be exclusive to Ver. 15.61 which apparently no longer has the 500 Mb governor on the drafiles folder. Although it wasn't unnecessary, we repeated this process in the backup user profile. After making this change and rebooting our computer, we have found that we still have to re-launch Dragon and usually reboot our computer but we can now go 4 days rather than being limited to 4 hours. Of course your mileage may vary but please report back. Debunking our little experiment is just as important as verifying it. If we screwed up, we will just blame the Big Haired Lady's Thunderbird punch anyway

PS: The Big Haired Lady is planning on supporting this 4th with live ammunition, mortar rounds, hand grenades and a 240 mm howitzer which she swears was purchased on eBay



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

Change "No" to "Know" w/KnowBrainer 2020
Trial Downloads
Dragon/Sales@KnowBrainer.com 
(615) 884-4558 ex 1



 07/04/2021 01:58 PM
User is offline View Users Profile Print this message

Author Icon
David.P
Top-Tier Member

Posts: 617
Joined: 10/05/2006

 

I'd suspect that something else is/was broken in your case.

 

For what it's worth, I've never had any problems with Dragon's speed due to my 5 GB sized user files, and I don't see any performance gains now that I've slimmed them by 96% down to 200 MB.

 

Also, as already mentioned the other day, Dragon uses only microscopic amounts of the memory and CPU power that is common nowadays. Particularly, it never uses those huge amounts of RAM that you are mentioning, but only a (by today's standards) ridiculously small amount of a couple of hundred Megabytes.



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

Sennheiser MKH Mic
Visual & Acoustic Feedback + Automatic Mic Control



 04/12/2016 05:16 AM
User is offline View Users Profile Print this message

Author Icon
Stephan Kuepper
Top-Tier Member

Posts: 2164
Joined: 10/04/2006

These are the standard folders that you should retain.

Technically speaking, you need only the current folder in order to have a working profile. However, Dragon will create the other folders as you work, and it is always a good idea to have a backup folder in the profile so you can restore it easily, should need arise.

I don't think anyone has ever bothered to research what exactly the other folders are for, or which functions you lose when you delete them.

Finally, to keep the profile manageable, check the "Conserve disk space for portability" option in Tools - Options - Data.

HTH, Stephan

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

www.egs-vertrieb.de - www.spracherkennungscloud.de

 07/05/2021 05:49 AM
User is offline View Users Profile Print this message

Author Icon
Stephan Kuepper
Top-Tier Member

Posts: 2164
Joined: 10/04/2006

David,

we have recently come across a number of similar cases. When Dragon gets the slows, deleting DRA files is the first thing we suggest. We suspect that Dragon is so busy writing DRA files that it doesn't have time for recognition any more. It is certainly not an issue of sheer profile size - I've had massive profiles, too, and never had speed issues.

One big difference is that many end users in Germany have Roaming profiles (in the U.S., this feature is rarely heard of). Sometimes it helps to turn the Roaming user off. Still I suspect that the culprit remains the same - Dragon writing DRA files when it's not supposed to, in this case, to the server.

My 2 eurocents, Stephan

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

www.egs-vertrieb.de - www.spracherkennungscloud.de

 07/05/2021 06:04 AM
User is offline View Users Profile Print this message

Author Icon
David.P
Top-Tier Member

Posts: 617
Joined: 10/05/2006

Stefan,

 

I concur that, while redundant files just sitting there usually don't slow anything down except possibly antivirus and search indexing software, trying to write files over a slow connection can actually drag things along.

 

Regards

David



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

Sennheiser MKH Mic
Visual & Acoustic Feedback + Automatic Mic Control



KnowBrainer Speech Recognition » Dragon Speech Recognition » Excessively large User Profile

Statistics
32107 users are registered to the KnowBrainer Speech Recognition forum.
There are currently 0 users logged in.
The most users ever online was 12124 on 09/09/2020 at 04:59 AM.
There are currently 267 guests browsing this forum, which makes a total of 267 users using this forum.

FuseTalk Standard Edition v4.0 - © 1999-2021 FuseTalk™ Inc. All rights reserved.