I have noticed that when I run the accuracy tuning it runs for about 30-40 min.  If I try to run it again (even 5 min later) it seems to go for another 35-40 min as though it did not work on the first occasion. 

 Looking through my speech file I have found that there is a folder called dataarchive.  With in that folder there are 300 other folders called session1 through two session300.  I was wondering whether these are records of my voice recordings and whether they should have been processed and deleted in accuracy tuning.  I wonder whether this is the problem.

 Could someone please tell me whether there should be 300 folders in the dataarchive folder and whether that is normal or if there is something going wrong with my set up?

Thank you

Simon

Simon,

you're KnowBrainer profile shows that you're using DNS 10 Medical. Are you using 10.0 Medical or DNS 10.1 Medical? It's not clear from your KnowBrainer profile.

As far as I can remember, the only reason why there should be anything in the dataarchive folder is if you're using Roaming user profiles. Otherwise it should be empty. If you're not using Roaming users, then I would simply delete all those session folders. From my recollection, it safe to delete them anyway.

The Acoustic and Language Model Optimizer will reevaluate any archive entries in your user profile, as well as any dra files that are located in your drafiles folder (\current\voice_container\drafiles folder). The fact that running the Acoustic and Language Model Optimizer a second time takes just the same amount of time as it did previously does not mean that the Acoustic and Language Model Optimizer didn't run or didn't correctly analyze and update your user profile. The fact that the Acoustic and Language Model Optimizer takes approximately the same amount of time the second time you run it is simply an indication of the fact that it's re-analyzing any dra files or accumulated (past or current sessions) archive data.

Lastly, take a look at the data contained in all of those dataarchive session folders as well as the dates. I would be willing to bet that there is some very old data there. Regardless, the Acoustic and Language Model Optimizer will use that information over and over again every time you run it.

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

"About the time we can make the ends meet, somebody moves the ends." Herbert Hoover

Thank you Chuck,

Your reply is very helpful.  It is good to know that I can delete the dataarchive folder.

My knowbrainer profile seems to be incorrect.  I am using DNS pro 11.5.  I have never used the medical version of DNS.

I have found the drafiles folder that you referred to.     It contains 354 files.   I am just wondering if there should there still be files present after accuracy tuning? 

I have just bought a new computer for work.  It is a dell i7 with 4gb ram.  I plan to increase the memory to either 8gb or 16gb.  Is there a noticable difference in DNS performance between  8 or 16gb ram?

 Also when I set up my new compter, do i just copy my user profile accross?

Thanks again

Simon

Simon,

this topic has been discussed before, so I suggest that you search the forum should you need more detailed advice.

Acoustic optimization converts .dra files into .nwv files that have roughly the same size, and yes, I agree that they should be deleted by the optimizer. You may safely delete all .dra and .nwv files that are stored somewhere in your user profile (the actual folder depends on your setup). All you lose is the ability to optimize your profile on these files. 

In my experience, acoustic optimization does yield tangible results in two cases: 1. on relatively new profiles (< 1 month), 2. in case of a slight loss of accuracy with an older profile. Other users may have other experiences. Anyway, we're talking about perceived accuracy here; no one has ever made any measurements. That's why I don't bother much about optimizing. 

Re new computer: the safest way is to export and import your user files. If you have a Roaming user set up, this is the only safe way to go; otherwise, you could copy the files into the appropriate folder (C:\ProgramData\Nuance\NaturallySpeaking11\Users on Windows 7, 64-bit).

Stephan

www.egs-vertrieb.de - Speech Recognition Blog - Forum: www.immer-eine-Nuance-besser.de

Here in 2008, first mention of a method how to make accuracy measurements, and some results (comparison between brand new user files, and user files that had been used and tuned over a period of about a year, the latter being +109% more accurate):

http://www.knowbrainer.com/PubForum/index.cfm?page=viewForumTopic&topicId=6469

Best Regards
David.P

Core i5 @ Windows 7 64bit
Sennheiser MKH HF Capacitor Desktop Gun Mic @ M-AudioTransit USB
Fingerworks iGesture Multitouch Pad & 3DConnexion SpaceNavigator + Visual & Acoustic Feedback + Automatic Microphone Control
NaturallySpeaking installed in SATA RAM SSD Solid State RAM Hard Drive

Download NaturallySpeakometer and objectively measure the perfomance of your NaturallySpeaking installation, compare the impact of different hardware (microphones, processors, hard drives etc.) on your speech recognition system speed -- also compatible to Vista, Windows 7 and Windows Speech Recognition

[edits necessary for this forum not having a "preview your post" function]

David, 

thanks for reminding me. Did you run the accuracy scheduler over those files or is it just normal adaptation and correction? Also, since in May 2008 you must have used DNS 9, how does DNS 11 behave?

Best, Stephan

www.egs-vertrieb.de - Speech Recognition Blog - Forum: www.immer-eine-Nuance-besser.de

Stephan, yes I run the accuracy scheduler about once a week.

Thanks for asking about DNS 11. Actually it runs quite good after I finally managed to work around a few quirks (and after Nuance has fixed a few others in SP1).

I haven't actually done much accuracy measurements between DNS 11 and 10 (only one quick shot here).

Cheers David

Core i5 @ Windows 7 64bit
Sennheiser MKH HF Capacitor Desktop Gun Mic @ M-AudioTransit USB
Fingerworks iGesture Multitouch Pad & 3DConnexion SpaceNavigator + Visual & Acoustic Feedback + Automatic Microphone Control
NaturallySpeaking installed in SATA RAM SSD Solid State RAM Hard Drive

Download NaturallySpeakometer and objectively measure the perfomance of your NaturallySpeaking installation, compare the impact of different hardware (microphones, processors, hard drives etc.) on your speech recognition system speed -- also compatible to Vista, Windows 7 and Windows Speech Recognition

[edits necessary for this forum not having a "preview your post" function]

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

Simon,

After running the Acoustic and Language Model Optimizer, your dra files, or at least most of them, will be converted to files with the extensions nwv and dft. These are archive files.

However, it isn't absolutely essential that you retain these files. Because you have so many, I would suggest that you delete all the files in the drafiles folder. While the dra files in this folder are used by the Acoustic and Language Model Optimizer, having so many simply means that the Acoustic and Language Model Optimizer has not properly removed and/or converted many of these. You can safely remove all the files in this folder and then simply keep an eye on this folder after running the Acoustic and Language Model Optimizer the next time. When you do, the dra file should be converted to nwv and dft formats. If they are not (don't worry if there are a couple of dra files left over) in this format for the most part, then I would run a repair of DNS (Windows Control Panel | Programs and Features | locate and highlight Dragon NaturallySpeaking 11, then click on Repair at the top of the Programs and Features dialog. If running at Repair doesn't correct this problem, then you may have a corrupted user profile. There are many forms of user profile corruption. Some of them are fatal and some of them are very subtle (i.e., subtle corruption is no prevent you from using your user profile, but they can and cause functional problems and errors).

As far as transferring your user profile from your old computer to your new one, I would strongly advise you to use the Manage User Profiles... | Advanced button | Export/Import features. On your older system, export your user profile(s) to a USB thumb drive (flash drive) or to an external hard drive. A USB thumb drive is the easiest to use, but I would recommend that you ensure that you have at least a 4 GB USB thumb drive (flash drive). Often you can get away with 2 GB, but 4 GB gives you more than enough room, especially if you have more than one user profile. Even so, why run the risk of running out of storage space. Stick to a minimum USB thumb drive 04 GB or greater. When you take the thumb drive to your new system, after installing DNS 11.5, simply reverse the process using the Manage User Profiles...

Also, if you're no longer going to be using DNS 11.5 on your older system, I would recommend that you uninstall it being sure that you're connected to the Internet so that the Nuance activation server will restore your activation.

In addition, when you install a new copy of DNS 11.5, you will be prompted to create a new user. I recommend that you do so. You can minimize the creation process by selecting "skip training" when you get to that portion of the new user creation process. This will give you a new user profile that is untrained, but you can always train it later if necessary. If you're using DNS 11.5 Professional, then you can always export your vocabulary using the Manage Vocabularies... option and imported into your new user profile by simply reversing the process again. This gives you your full vocabulary ported to your new user profile including all properties, training, and custom words. As far as your custom commands, you can simply copy the MyCmds.dat file from your imported user profile to your new user profile. This will give your new user profile a complete copy of your custom vocabulary and your custom commands as noted above. If you need to train your new user profile, you can do so at any time.

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

For the things we have to learn before we can do them, we learn by doing them. - Aristotle (384 BC - 322 BC)

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