I've heard over and over the import of good audio quality for accurate recognition rate.  Often the focus is on the microphone, then soundcard, and then computer if I understood the theory correctly.  I've tried using DNS on various tablets and desktops without ever having much success.  However, if I dictate into my Sony digital recorder using the same analog Plantronics Headset that I use on the computer, and then through the Sony Digital Voice Editor 3 software transfer the audio file to the computer and activate the DNS 10 voice recognition software I get substantially better accuracy.  This seems to hold true regardless of whether I train the profile or not, and whether I set the digital recorder quality to HQ or SHQ. 

When I use my headset on the computer the DNS 10 indicates that my audio quality is good.  I'm currently using the HP tx2-1370us tablet becuase it is faster than my Lenovo X60 and my desktop.  I disable the HMA!proVPN and the X1 search software when using DNS 10 without noticible gain.  

Aside from potential external line interference when I use the headset to dictate into the computer, don't both methods rely on the same hardware performance?  Is the exception the speciification of the audio file created?

 What can I do?  Configuration, different computer, stop drinking?  Please advice.  Thanks, DumbDog.

 

 

-------------------------
DNS Perferred ver 10.0 on OS: W7 Ultimate w SvcPak1, 64 Bit, TrendMicro Titanium ver 3.1.1, MS Office Pro 2007,  X1 Pro Client ver 6.7.3 (build4054dh) - search software, HMA!proVPN ver2.6.9.   This is a tablet pc - HP model: tx2-1370us /PN: VM307UA#ABA

Don't! Most likely, the problem is with your soundcard or anything else in your computer interfering which doesn't come into play when dictating to the recorder (although according to the "rules" you should get poor accuracy also due to its being substandard).

First thing, next to going on drinking, get yourself a USB headset to connect to the computer, if the one you have isn't such (which I assume, otherwise you couldn't connect it to the audio recorder).

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!

Rudiger,

Thanks for your input.  I apologize for the typo in spelling your name - don't know my ASCII codes well enough to put points above the "u".  I can't stand to type with all the writing I have to do.  When I see people commenting that they are at 100 words per minute is that reality for the masses after practice and training?  I'd be happy with 50 or 60 words per minute - and probably couldn't dictate with decent enought sentence structure and punctuation at 100 per min.

 

I've got a USB headset.  It's a middle of the road, or maybe less than lower part of the road, depending on the road from whence the came - Logitech headset that was in the $50 or $60 dollar range at Best Buy if recolection serves me correctly.   The reason that I ended up using the analog Plantronics aside was initially that I was able to obtain a better quality index number on DNS 10 using the analog set...    Yes, I do see the seemingly obvious logic in concluding that the the Logitech USB headset is a piece of junk given an assumption that one should be able to obtain better qaulity audio via the digital route on a computer.   But, when I tested the analog vs digital on my Lenovo x60 tablet the two were fairly equal - I assumed that the Lenovo running 32bit XP only capable of seeing 3gb ram would not run DNS as well as the HP.   My assumption with respect to HP tx2 tablet having poor usb audio quality was that the W7ultimat x64 OS did not have OEM hardware drivers deployed, whatever W7 found and installed is what it has.  Alternatively, I also have the original W7home OS32bit installed on it which has also had a truckload of HP craplets installed -  I attempted to test the USB headset with DNS on this load and ran into some configuration confilcts that most likely could have been easily resolved but I didn't want the hassle.

This brings me to my next question, which OS is best for DNS 10? :XP Pro 32bit or W7 Ult 32bit or W7 Ult 64bit given that I will max at 3gig Ram on XP.

When I decided to start trying to adapt to DNS I reviewed the DNS for computer recomendations and didn't find anything very helpful.  I'm coming into this world without any benchmark to know what is good and fast as opposed to as good as it gets with the DNS software.   I know the HP is faster than any of my other laptops - but I also consider it to be of lesser quality than the legacy lenovo x60 and than a fujitsu T1010 tablet that my kid is using - it came with Vista and I didn't want to hassle with config. issues on W7 or XP (Although I suspect it has the best Audio Card of the bunch - I had to look at it for the model # and I'm pretty sure it said Altec Lansing on it.  

So, I'd rather start with the proper computer before investing in a quality headset - my kid will off the headset in the blink of an eye.  What is most important when it comes to hardware?   Ram, Processor, Bus, Audio card, static HD?   Would I see a significant performance increase if I switch to a desktop?  If so, I'd be inclined to build one so I can avoid all the proprietary junk.  All I want my DNS computer to do is DNS -it could even stay off the internet and avoid running a process consuming firewall/antivirus software package.  

All things being equal - stripped of all excess software:  Would I experience a significant performance difference in DNS short of going to the latest greatest faster hardware?  I'm the type that would rather have the Toyota Camery of a computer than deal with memory and chip conflicts or what not.

 Please help this poor mutt.  Give the dog a bone.  thanks, Dumbdog.

 

-------------------------
DNS Perferred ver 10.0 on OS: W7 Ultimate w SvcPak1, 64 Bit, TrendMicro Titanium ver 3.1.1, MS Office Pro 2007,  X1 Pro Client ver 6.7.3 (build4054dh) - search software, HMA!proVPN ver2.6.9.   This is a tablet pc - HP model: tx2-1370us /PN: VM307UA#ABA

No problem. The ASCII code would be 0252 incidentally.

Don't worry about it, and don't get intimidated by such remarks. Better yet, try to get a start with it first according to what you have and what you can do with it, step by step. I would definitely suggest getting the USB headset working, it may not be the best under the sun, but it should work reasonably well least, minimally, if it is not broken somehow. If the problem is with your system not being able to provide the appropriate drivers, start from there, by looking into how to deal with such hardware issues. Make a recording using the Windows sound recorder in order to test it, and if it fails, try the same on a different computer to determine whether it is the computer or the headset.

As far as operating systems, it won't make so much difference whether you have Windows XP or Windows 7, as regards running Dragon, at least not at this stage where you are. Just get started, would be my advice, and take you time improving everything else as you go along.

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!

DumbDog,

Just a little tip.  To add a "u" with the points above as so ü

if you dictate your draft post in Microsoft Word first, if you then say, "Insert" to select the "Insert" tab at the top of the document, then say, "Symbol" (this will then drop-down the Symbol list).  Then you say, "More Symbols."  This will then bring up the Symbol window and you will see the ü in there.  You then just highlight it and say, "Insert" to put it into your Microsoft Word document.  When your draft is finished in Microsoft Word, you then just copy & paste it into here. 

I'm not sure if there is a Dragon command to bring up the symbol window quicker when in Microsoft Word but I'm sure if there is, someone will inform us so.  If you had KnowBrainer, when in Word, you could then simply say "Symbol" and it would immediately bring up the symbol window with your ü in there.

What I have done as I suspect a lot of other people in here may have, is to add Rüdiger to the vocabulary or made a simple text macro for his name.  Then it should come out on screen correctly every time you say it :-).

Cheers

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

Tiger Feet

| DNS 12 Professional | KnowBrainer 2012 | Windows 7 Professional /64 Bit | Intel Core i7 Quad Core 3840QM (2.80GHz) 8MB | 16GB RAM. | 240GB Intel 520 Series SSD Boot Drive | 750GB 7200 rpm Secondary Drive  | Plantronics Wireless Handsfree Microphone CS60-USB. |

Another way of achieving this is to say "type u umlaut".

R u umlaut diger

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

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!

Sorry, I have to ask.

 Are you at the exact same position by your computer when you use your digital recorder as when you dictate through directly to your computer?

 I would tend to be closer to my computer when I was actually using it then when I wasn't, so if there was some noise source around your computer that your direct dictation was picking up...

 You haven't actually said that you are even in the same room as your computer when you use the digital recorder.

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

----

 .

 .

 .

 .

Help! I can't see vowels!! (and sometimes "y")

Rüdiger,

 I have not been positioned consistently for session to session when using the computer to dictate.  The USB head set has been trashed by a 14 year old but works -  it appears that the kid hung it from a lamp as one side is partially melted.  I will replace it ....

I use the sony digital record as my mobile device and rarely within consistent environmental variables.  In fact I do not always us the head set.

I do have a seperate profile for each device, i.e. Sony Digital Recorder, HP with Analog Headset, HP with USB Headset, as well as unique profiles on the Lenovo X60.  I have made a couple observations:

1.  The Sony Digital Recorder has some noise filterig functions as well as abilty to select codec and quality variables.  It is currently set to SHQ quality and using a codec that produces a .wav file with 44.1kHZ.

2. When files are transfered from the Sony recorder via USB cable to the Sony Dig. Voice Editor sorftare the files are set to convert into mp3 files; then they get transcribed if desired.  I assume that the quality of audio file is bettter on the Sony D.R. than the computer; and that the computer may not be able to compete on that topic.

3. The HP Tx2-1370us tablet has a fan that is frequently running on a balanced power profile and that is constantly running on highpower -  I have two of these systems, one failed early on and the MB was replaced under warranty; it runs better but still throws off constant heat and fan still running a lot.  Both of these units have noise producing fans.  Given the assumption that the hardware performance should not create a massive baseline performance variance in DNS; I'm inclined to dump the HP on my kid and take the fujtsu.  I'll check audio hardware specs and see if driver availabilty or compatabilty is better with W7 or XP and load it with whatever is most suitible.  I'm going to scrap the tablet functionality and configure it without all the garbage services and hardware extras.  I will keep the system isolated and offline with minimal software... DNS10, MS Word-07, and oem audio package. 

4.  If consistent position of user, computer, and mike is critical, the I think it would be best for me to get a mike that sits stationary on a boom.  The room I am working out of is pretty quite, but...  do I need to concern myself with electrical noise?  Should I throw a power conditioner on inbetween the wall and the computer?  What about overhead can lights with those flourescent effeciency bulbs - any potential frequency problems?  What about 5.8gHZ wireless phone transmissions? 

5.  How far up the quality scale should I go with the purchase of a boom mic? 

6.  Starting with a fresh profile after I've played the the audio recording quality via windows and my pyschical positioning - what would you recommend I do to prepare the profile?  Less or more, harder or easier, or maybe none at all initially?

7. Once I begin to dictate....   Are there some policies or pratice which I should adhere to strictly such as:  if a mistake occurs in a sentence should it be correct immediately -  on the spot, on the fly?  or can I come back once I have completed my draft?  I presume that I should configure DNS to retain the biggest training files possible? or  ????

I've got tons of time invested in attempting to make this happen - started attempting make the transition from my keyboard in earnest Q4 of 2009 and have never had much success making it really sing: I have spent a lot of time reading those training stories to it.    I suspect consistiency in operation is critical, or shoudl the DNS software just intuitively figure me out as we move forward?

 Here goes dumbdog to play computer techie and load a system...  ? perhaps I have a clean Acronis image (.tib) already to be burned on the drive???  that would be too easy.

 thankyou for all the advice, best regards from dumbdog!

-------------------------
DNS Perferred ver 10.0 on OS: W7 Ultimate w SvcPak1, 64 Bit, TrendMicro Titanium ver 3.1.1, MS Office Pro 2007,  X1 Pro Client ver 6.7.3 (build4054dh) - search software, HMA!proVPN ver2.6.9.   This is a tablet pc - HP model: tx2-1370us /PN: VM307UA#ABA

Quote:

Logitech headset that was in the $50 or $60 dollar range at Best Buy

Believe it or not, the Sony Digital recorder is decent but not optimally suited to speech recognition. However, even premium digital recorders should always produce lower accuracy than microphone input because you are not going through a middleman, so to speak. A lot of end users are getting by with Logitech microphones but they're not good enough to make our Microphone Comparison Matrix. We suspect that if you switched to a professional microphone, like the Audio-Technica Pro 8HEmW with a USB soundcard, you'll be tempted to crush the Logitech microphone under your heel. Now onto your questions…

1. We don't recommend using any software based or hardware setting noise filtering functions on any digital recorder. The only noise filtering you should employ is Dictation Mode (as opposed to Conference Mode).

2. Think of your digital recorder as a combination microphone/USB soundcard that saves recordings of your dictation. MP3 files are fine but if your digital recordings are more accurate than your microphone, your speech recognition microphone is substandard because it should be the other way around.

4. The proximity range of the condenser microphone that is built into digital recorders is somewhat flexible and quite decent. Most speakers are not good at locking a precise distance from the microphone but the microphone is reasonably flexible and recorders include auto level input control to give you even more flexibility. However, consumer grade speech recognition microphones typically include small microphone elements which have very little tolerance. Small microphone elements only have about .5 inches of sweet spot tolerance before you have to rerun the Audio Check. Larger diagram professional microphones are much more tolerant of a little sloppiness which is another reason to avoid Plantronics and Logitech.

You'll find some speech recognition resellers recommending that you run the Audio Setup Wizard every time you launch NaturallySpeaking while others, such as everyone in our company, typically only recommend running the ASW when we swap out the microphone. We have often wondered why so many speech recognition resellers recommend rerunning the ASW every time you launch NaturallySpeaking and eventually figured out that they were using the manufacturer OEM microphone or something equivalent.

5. Are you referring to theBoom line of microphones? If so, they are all identical and ideally suited to speech recognition but they are no longer the Gold Standard. In our opinion, the new Gold Standard is the previously mentioned Pro 8HEmW which costs less than most of theBoom models.

6. Forget about training. Training not only offers you the least bang for the buck but can often reduce your accuracy rather than help. Unless you are suffering from substandard accuracy, we don't recommend more than a single training session and when we upgraded to Ver. 11, we opted out of training which brings us to our next recommendation. If speech recognition plays a major role in your work environment, you should consider upgrading to Ver. 11 which is considerably more accurate with digital recorders, includes a doubled sampling rate and a host of other features you'll find in our NaturallySpeaking 11.5 Review

7. Many end users worry far too much about mistakes. NaturallySpeaking's "Silent Adapt" feature is pretty good at studying work habits but mistakes have almost no impact on your accuracy. Don't worry about them. Although good dictation habits are the # 1 item to focus on, our favorite utility is the DNS Vocabulary Editor. Although training an occasional misrecognized word can help, most of the time we prefer creating Vocabulary Editor rules which we have covered in depth in our new KnowBrainer Training Guide. Bottom line: NaturallySpeaking's actually pretty good at paying attention to your work habits but it does need a little help along the way.

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

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

 

 

 

Rüdiger:

#6) Training:  (after hardware testing)  DNS10 - preferred.  I preformed a fresh OS install on a Fujitsu T1010 Tablet.  Tested that tablet against my 2 other tablets: HP Tx2-W7Ultx64 / Lenovo X60-XPproSP3 and compared the Logitech USB and Plantronics Analog headsets.

Best preformance is Lenovo X60.   I perfomed a minimun training.  I disabled all of the filtering on the x60 SoundBlaster properties. I set the Sony DVR to low-Mic and Super HQ mp3 format.  I ran the "vocabulary optimizer".  I selected approximately 10 letters and 5 emails (all lengthy and representative of currenty style/needs).  I then went through the vocabulary and added appropriate unrecognized words.  I found my accuracy to increase significantly after configuring vocabulary words.  Accuracey appeared to increase again after correcting several dictated files and additional vocabulary config.  I then added a "mic-in" source to the profile; dictating directly into Dragon Pad worked well.

BUT...  Accuracy appears to have taken a noise dive after I dictated several documents and emails that were not corrected.  Upon restarting system with intent to correct such, I found that the DragonPad upon opening any of these uncorrected rtf's no longer had any audio linked to it, nor did I locate dra files.  

In your last reply; "...Training not only offers you the least bang for the buck but can often reduce your accuracy rather than help..." 
[It appears that 'training' with uncorrected docs. or otherwise reduced my accuracy.]

Q6.1: Is preforming corrections important?  [user is set to 'save' on exit; optimumization is scheduled to run daily.]

[I tend to work in 20hr+-/3day stretches and in reviewing the audio files there is a noticeable difference in my speach after day 1 that gets progressively less crisp and less fluild over day 2 and 3]  

Q6.2:  Is it likely that my reduced accuracy is caused primarily by using a single profile for dictation while sharp, crisp, and fluild; and while impaired?  If so, would I need to setup an additional 'diction source' (like "impaired" for my user profile? 

Q6.3:  DNS11.5 Preferred upgrade is a good chunk of change ($140 I think) and I hate to dump more $ on voice recognition without testing whatever possible solutions for cost-effectiveness. 
Q6.4:  Is there a trial version available? 
Q6.5:  If not, does the $50 "Home' version have the same improved accuracy for DVR's as the Preffered version?  
Q6.6:  The top of the line Sony DVR - ICD-SX700D  ($149) indicates "Featuring Dragon NaturallySpeaking® Voice to Print Speech Recognition software"; Is this included in DNS11.5 Home?  In DNS11.5 Pref? 

[It seems apparent that using a DVR would be best for me. If the "Voice to Print" is a uniques module perhaps I would be best to purchase the new DVR.]

Please advice.  Thanks, DumbDog!

-------------------------
DNS Perferred ver 10.0 on OS: W7 Ultimate w SvcPak1, 64 Bit, TrendMicro Titanium ver 3.1.1, MS Office Pro 2007,  X1 Pro Client ver 6.7.3 (build4054dh) - search software, HMA!proVPN ver2.6.9.   This is a tablet pc - HP model: tx2-1370us /PN: VM307UA#ABA

We would like to apologize in advance for the following but hopefully you will find some value in our response:

It would appear that you need speech recognition for professional work and you're trying to skate by on consumer grade software and hardware which means you're kind of getting what you paid for. If you're worried about the price of the DNS Premium upgrade, that's likely to be a problem because if speech recognition is important to your workflow, you have to buy the right tools for your job. You'll never see a mechanic buying knuckle busting used Kmart wrenches on eBay but that's exactly what you're attempting.

The reason why your recordings disappeared after you closed NaturallySpeaking is because you are using a consumer version of NaturallySpeaking that does not include professional amenities such as Deferred Correction. You would have to upgrade to the $395 Preferred/Premium to Pro 11.5 Upgrade in order to link recordings to your documents so that you could listen to those recordings at a later date. Nuance considers Deferred Correction to be a professional feature and that's why the price of DNS Pro is 3 times the price of DNS Premium even though it utilizes the same speech engine.

That so-called top-of-the-line Sony ICD-SX700D recorder is actually barely acceptable. These types of consumer grade recorders typically produce 50% more errors and don't include professional amenities like seamless glitchless punch in, cue and review and a host of other professional features. In other words it's going to take twice as long to get the job done. The only digital recorders that we recommend are models that incorporate the Olympus DS2 12 times compressed lossless fidelity algorithms but those recorders start at $249 for the consumer grade model and go up to $450. The Home Edition of NaturallySpeaking cannot be used with a digital recorder. We think of the Home Edition as more of a speech recognition appetizer or stripped-down version of DNS Premium.

There are no trial versions of NaturallySpeaking although Nuance includes a 30 day customer satisfaction moneyback guarantee.

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

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

 

 

 

Lunis Orcutt

"I watched a snail crawl along the edge of a straight razor.  That's my dream.  It's my nightmare.  Crawling, slithering, along the edge of a straight razor ... and surviving."That is a quote from the 1979 film Apocalypse Now; Brando as Colonel Walter E. Kurtz.  However, I find that the quote has been applicable to warfare in the IT jungle: My bottom line shifts from black to red every time I find myself on the bleeding edge of any technology.  Secular market trends, like voice-recognition, will invariably lead to a paradigm shift forever changing the balance of ownership from: technology to man - to Mankind owning technology through the "Voice-Intelligible Interface":  so easy-to-use that even a "DumbDog” could figure it out.Every 3 years am afflicted with "DumbDog Denial":In 2006 with Sage's software: ACT! 2006 ver. 8.
In 2009 with Acronis's software: Advanced Server/Workstation ver.10
In 2012 with Nuance’s software: Dragon Naturally Speaking?  No thanks!My Uber conservative SEC attorney bestowed upon me the wisdom that is "DumbDog, if you have to ask-you already know the answer.”I've been dabbling in voice-recognition since the beginning in fact back in the year 2000 one of the last public plays I was involved in was in IT Training entity, and the mascot and namesake was very effectively utilizing ViaVoice or Dragon to write textbooks.  Of course that Nut-Job was/is a crazy SOB that to this day is still "Crawling, slithering, along the edge of a straight razor … and surviving.”In the 4th quarter of 2009 I led my 3 man team in my conquest to deploy laptop’s configured with Acronis, connected to our servers, as functioning IT control/support model “proof of concept” – Beta Test deployed to sales personnel within our distributor channel.  So it was DumbDog leading a pack of 2 "so smart they’re stupid" engineers.  The secondary requirement in the grand-plan I had was a bet that speech recognition would become mainstream by 2013-2014; such that it would be more cost-effective and efficient to migrate a computer illiterate worldwide network of distributors to computerization and full-on technology integration and savvy, by teaching them a standardized set of voice commands to control and operate their computer instead of attempting to force feed them a keyboard and mouse.  After 4 months, 24 licenses, 5 laptops, and 3 servers, I finally arrived at the truth of the matter: product was not ready for prime-time.  Of course there are no refunds for software or for time and money wasted:  lessons learned are valuable-this time around it didn't affect our business operations OR run into the tens of thousands as was the case in 2006.  Ironically both Sage and Acronis experienced difficulties related to SQL (and their recent adoption of it).No apology is necessary from YOU!  There is much greater value in your response than you realize.  I am in agreement with your view that it is necessary to have the right tools for the job.  On this note, my initial inclination when I began digging back into this last month was to start with defining a computer build, OS, and peripheral hardware.  What I learned was that there is a difference of opinion amongst the savvy elite as to best practices with hardware build, software configuration, and operation to achieve optimal results.It was only after a couple hours of beating my head against the wall searching for clarification on “DRA” files- to no-avail that I searched my data archives.  Can you please point me to the location of the data such as DRA files require the Professional License version?  It was not until I reviewed my notes from 2010 that I realized I had encountered this same exact problem!  Logic dictates that this is DumbDog being dumb – surely there is a key phrase that can be entered into the search field and “presto” – it will return such simplistic details.As an aside, is there not any scientific Empirical data that can be referenced with regard to the effects on accuracy and performance from microphone the microphone, computer to computer or OS?  While I don't claim to be any kind of expert on voice recognition software; I do have some expertise in sound propagation, specifically outdoor voice intelligibility. In this line of work the lives of the Armed Forces and 1st Responders are on the line-we use digital instrumentation to measure performance. The only area that I find myself in disagreement with you is your contention that I am "kind of getting what you paid for."  To that contention good sir; I must strongly refute.  I'll bet you a 6 pack that prior to ordering version 9 and version 10 that I had detailed communications with the good folks at Nuance; or at least that I had sent detailed comms.Right now I can get my voice to test generated at $0.0125 per word through Speak-Write.  At this juncture I need to get a handle of alleged performance enhancements between v11 packages and v10.  Please point me in the correct direction to ascertain the critical details on software versions.Thanks, DumbDog!

-------------------------
DNS Perferred ver 10.0 on OS: W7 Ultimate w SvcPak1, 64 Bit, TrendMicro Titanium ver 3.1.1, MS Office Pro 2007,  X1 Pro Client ver 6.7.3 (build4054dh) - search software, HMA!proVPN ver2.6.9.   This is a tablet pc - HP model: tx2-1370us /PN: VM307UA#ABA