![]() |
KnowBrainer Speech Recognition | ![]() |


|
Topic Title: DNS Medical 10.1 vs DNS Pro 11 + Med Vocab Topic Summary: Which is better while we wait for DNS Medical 11? Created On: 01/15/2011 04:43 PM Status: Post and Reply |
|
![]() |
- talkdoc | - 01/15/2011 04:43 PM |
![]() |
- Lunis Orcutt | - 01/15/2011 10:33 PM |
![]() |
- mcampbell | - 01/19/2011 07:46 AM |
![]() |
- talkdoc | - 01/19/2011 09:05 AM |
![]() |
- Lunis Orcutt | - 01/19/2011 12:08 PM |
![]() |
- Chucker | - 01/19/2011 01:49 PM |
![]() |
- talkdoc | - 01/19/2011 10:56 PM |
![]() |
- mcampbell | - 01/19/2011 01:36 PM |
|
|
|
|
I'm wondering what you would recommend for someone with DNS Pro 10.1 who would like to upgrade to Medical, given that Medical 11 has an unknown due date.
------------------------- |
|
|
|
|
|
|
|
|
You could upgrade to the Dragon Medical 10.1 Upgrade but it still utilizes the Ver. 10 speech engine. The general consensus is that Medical 11 won't be released until around July and that upgrade will probably cost considerably more than the current Medical upgrade but you have another option. You could upgrade to the NaturallySpeaking Pro 11 Upgrade and add the KnowBrainer specialty vocabularies which would be considerably more accurate, not to mention less expensive, than the current version of Medical
-------------------------
|
|
|
|
|
|
|
|
|
that's what I'm doing right now and the result is very satisfying. Accuracy is markedly enhanced by running a lot of prior notes through the Accuracy Center. For a couple of weeks I selected old notes from everyone I was going to see that day and the voice profile improved significantly. I saved the most time and not having to correct "nonmedical" text such as parts of the HPI 1 of the things that would describe what the patient was doing. Overall, it's been well worth the cost
|
|
|
|
|
|
|
|
|
Thank you both. Lunis--I've gone ahead and ordered DNS Pro 11 + Med Vocab from you on the strength of your recommendation. But I'm still curious--the gist of prior discussions is that Medical is in a class by itself, engineered differently, so to speak, and not 'just' Pro + specialized vocab/language models. If I understood those discussions rightly (a big if), I get it that 11 is such a big improvement over 10.1 that the equation has changed (DNS Pro 11 +Vocab > Med 10.1). If you get a chance, though, I'd love to hear your thoughts about what difference Medical makes. mcampbell--Very interesting that accuracy improves with running notes through the Accuracy Center. Again, someone just listening to the discussions on ver. 11 would be struck by 'accuracy improved right out of the box without training,' and might presume that the addition of medical vocabulary would prempt the need for running notes. Good to know. ------------------------- |
|
|
|
|
|
|
|
|
Although we strongly suspect that Medical 11 is going to include a number of new features (including contributions from IBM R&D), from what our clients have told us and our own experience with previous versions of Medical (versions 4 through 10.1), we have seen absolutely no difference in accuracy when comparing Preferred/Premium or Professional with a professional third-party language model (vocabulary on steroids). We have heard a number of statements, from Nuance, about the algorithms being tweaked specifically for the medical profession but we've never seen any evidence of higher accuracy and a lot of our clients have even claimed better accuracy using a third-party vocabulary over the Nuance vocabularies but we suspect that both products are evenly matched (from an accuracy point of view). Please keep in mind that even though the current version of Medical, is now out of date (from an accuracy point of view), it still includes some unique features such as the hidden text box, a more user-friendly Command Browser and is the only DNS product that can fully utilize the Nuance Dictaphone microphone.
Medical 11, rumored to be released this July, may be a game changer but for the time being, from an accuracy point of view again, we believe the combination of a third-party vocabulary and NaturallySpeaking Pro 11 is the best way to go unless you're working with an EMR application and even then, keep in mind that our Medical Vocabularies also now include KnowBrainer 2011 which currently includes a workaround for dictating into EMR apps and will hopefully soon include the ability to seamlessly dictate directly into any EMR application. We have never been very pleased about Nuance crippling nonmedical versions of NaturallySpeaking to prevent them from working within EMR applications and have doubts about the legality of that practice. -------------------------
|
|
|
|
|
|
|
|
|
Quote: Although we strongly suspect that Medical 11 is going to include a number of new features (including contributions from IBM R&D), from what our clients have told us and our own experience with previous versions of Medical (versions 4 through 10.1), we have seen absolutely no difference in accuracy when comparing Preferred/Premium or Professional with a professional third-party language model (vocabulary on steroids). We have heard a number of statements, from Nuance, about the algorithms being tweaked specifically for the medical profession but we've never seen any evidence of higher accuracy and a lot of our clients have even claimed better accuracy using a third-party vocabulary over the Nuance vocabularies but we suspect that both products are evenly matched (from an accuracy point of view). Please keep in mind that even though the current version of Medical, is now out of date (from an accuracy point of view), it still includes some unique features such as the hidden text box, a more user-friendly Command Browser and is the only DNS product that can fully utilize the Nuance Dictaphone microphone. Lunis et al., First of all, the rules of the game have changed. Beginning with DNS 11, the EULA (End-User License Agreement) prohibits third-party vocabularies containing vocabulary that includes enhanced Language Models for versions of Dragon NaturallySpeaking below the Professional versions (Legal, Professional, or Medical). That is, it is perfectly permissible to distribute third-party vocabularies, particularly specialty vocabularies, such as medical, specialties or legal vocabularies, for DNS Home or Premium (DNS 11) that do not incorporate any related Language Models. This is why Voctool.exe has been eliminated from DNS 11. You can't talk about accuracy relative to any version of DNS if you're simply going to talk about the General-Large vocabulary. There are no differences in terms of accuracy among any of the versions of DNS relative to the general vocabulary. There never has been from DNS 4.0 right on up through DNS 11. The Language Model in all versions for the General-Large vocabulary is identical. Therefore, talking about differences in accuracy among the various flavors of DNS (Home, Premium, Professional, Legal, or Medical) is a complete non sequitur. Accuracy is the same. The only difference as regards accuracy applies to the specialty vocabularies in those versions of DNS (Legal, or Medical) that contain such because these specialty vocabularies also include a related specialty Language Model. The only thing that these Language Models to is improve the accuracy of DNS relative to the specialty vocabularies. If you want to see the difference, try using DNS 10 Medical with the Cardiology medical vocabulary and language model and try to dictate a standard e-mail. This is simply because these Language Models are not designed for standard dictation. They are designed specifically for the specialty Incorporated into specific specialty vocabularies. Quote: Medical 11, rumored to be released this July, may be a game changer but for the time being, from an accuracy point of view again, we believe the combination of a third-party vocabulary and NaturallySpeaking Pro 11 is the best way to go unless you're working with an EMR application and even then, keep in mind that our Medical Vocabularies also now include KnowBrainer 2011 which currently includes a workaround for dictating into EMR apps and will hopefully soon include the ability to seamlessly dictate directly into any EMR application. We have never been very pleased about Nuance crippling nonmedical versions of NaturallySpeaking to prevent them from working within EMR applications and have doubts about the legality of that practice. I generally concur with Lunis' assessment with regard to overall accuracy and third-party vocabularies sans Language Models specifically based on such, but this again is kind of a non sequitur. Any Professional version of DNS allows the user to create their own "Language Model" even in the absence of the Voctool.exe in DNS 11 versions because you can still adapt your user to your writing style by analyzing documents, which creates additions to the User slot that are just as effective as if you were able to use Voctool.exe and use the Middle slot. Not having accessibility to the Middle slot is not a significant disadvantage when using the current "Learn from specific documents" in the DNS 11 Accuracy Center. One of the reasons why this was eliminated in DNS 11, aside from preventing third-party developers from creating vocabulary/Language Model (Middle slot) vocabularies, the Middle slot was abused in most cases because it simply wasn't necessary, and still is currently unnecessary, and the way it was being used didn't really add any significant degree of improved accuracy even with specialty vocabularies. The User slot is sufficient for 99% of all cases. So, if you import a collection of specially vocabulary words, you can always create your own adaptation to the current Language Model by using the "Learn from specific documents". Accuracy will improve for specialty circumstances, such as filling out patient records yada yada yada where specific vocabularies are required by using this method. For the most part, they will be just as accurate as if you were using DNS Medical with both the specialty vocabularies and the associated Language Model. So, there is no reason to be overly concerned because the differences in accuracy are inconsequential and often not noticeable. Any arguments to the contrary are simply moot. Lastly, there has been a recent case with regard to the legality of the practice that Nuance employs with regard to EMR's. The case was thrown out of court because the judge determined: (a) there is nothing that legally prevents Nuance from doing this, and (b) the lawsuit was considered frivolous by the presiding judge in this case. The way that I would put it is that, regardless of any question of legality, it is simply not good PR and perhaps not ethical. Be careful about discussing legalities because that is something that can only be determined in a court of law, not the court of public opinion. The court of public opinion does not establish legal precedent. On the other hand, I do know that at least some of the reasons for Nuance's approach are reasonable and legitimate. This is one point that was made with regard to the written opinion of the judge in the above case. That is, the plaintiffs did not make a valid case relative to the explanations provided by Nuance.Chuck Runquist Be careful what you wish for. You may find that what you get is not what you expect, or what you want. - Aesop (620 BC - 700 BC modern interpretation) -------------------------
|
|
|
|
|
|
|
|
|
Thank you all--very helpful.
------------------------- |
|
|
|
|
|
|
|
|
I found a definite difference however one of the things that might have influenced this is not I "prepped the charts", much as the nurses prep a physical chart. The notes that I ran through the Accuracy Center were notes on the same patients that I was seeing that day. So the same problems in medications were coming up and probably the style of dictating was very similar. It didn't take much time and I think that the application could anticipate what I was going to say with greater ease
|
|
|
|
|
FuseTalk Standard Edition v4.0 - © 1999-2013 FuseTalk™ Inc. All rights reserved.