Home Forums Discussions General Discussion Voice Recognition

Anyone using voice recogniton programs with AC, any experience with Dragon Naturally Speaking?

J:

I am about to go "live" with Amazing Charts (AC) next month. In anticipation of that, I have been fiddling with Dragon Naturally Speaking v.9 (DNS) for a couple of months. I got the Preferred version to try out (only about $250 vs. $1000 for the Medical version). I have generated dozens of AC records, and dictated numerous letters using the program.

I will first say that DNS has improved *significantly* from the old versions (I would never have bought it in the past). There are still occasional glitches, but I'm not ready to say that they might not be solved by a better mic (I'm using the one that came with it), more training, or in certain cases, by springing for the Medical version (more about that later).

As far as using it with Amazing Charts, it works very well. Now, I have found that I have to open DNS several minutes before opening AC and let it completely "warm up", then open up AC, or it's more glitchy somehow. However, once they're both open, they work great together. I have a friend who has been using them together for months and is very happy with the set-up (this is the reason I sprung for DNS). My friend also uses the Preferred version, and says he has had very little trouble out of it. My friend may be a little more casual in his dictation style than I am (he may use fewer big words and uncommon medical terms).

With the Preferred version of DNS, I have had trouble getting the program to recognize some medical terms such as "auscultation" and "HgbA1c" and "Vytorin", despite the fact that I let DNS analyze about a year's worth of old transcription files (they were fortunately in .doc format and available to me) before I ever started. I had thought that scanning those old notes would populate the program's vocabulary with enough terms for my purposes.

However, it turns out that even if DNS has encountered the medical term in your old notes, it still has to make a verbal map of the word and match it to the text version, so it can type the right word when you say it.

DNS also apparently has the ability to recognize context -- this word is usually found next to that one, etc. and there is no context available for these uncommon terms in my version (although it apparently "learns" context, if you keep using that term long enough).

I am sure the Medical version has a tremendous medical vocabulary, and more verbal maps of medical terms. I assume there is more context information, also. Enough to lay down the extra grand? I don't know; I will have to play with the cheapie version a while longer, for the following reason:

It remains to be seen how much time I will actually spend dictating into AC with DNS. The templates feature will eliminate some dictation, especially on examination. I will probably only use it to round out the HPI after my nurse starts it, for unusual physical findings, and for discussion under "Plan". Actually, I have been thinking about trying out the "Instant Medical History" interface, and that might reduce dictation even more. I may end up with so little dictation as to keep glitches pretty rare (unlike when you're dictating your whole note from scratch).

Hope this helps,
Brian Cotner, M.D.

Using Via Voice medical edition (which is a lot less expensive than DNS). Have used it for three years satisfactorily. Occasionally need to turn off ViaVoice and restart when memory usage becomes too large.

I am using Dragon preferred also. However, at this time, I am less than thrilled.

I have trained straight in both at home and at the office multiple times. Yet I still get errors frequently. Often these errors occur, on terms that I have previously trained Dragon on. There are times the program works flawlessly, and times it doesn't seem to realize him speaking English. At times like that, it's just not worth using the program.

I now include a disclaimer at the end of all of my notes were accused Dragon, as is not as reliable as I would like, I don't always catch the errors.

It will definitely try turning on Dragon several minutes before amazing charts. Perhaps that's part of the problem, as it tend to turn on everything at one time.

Disclaimer: This post was created by voice recognition software and may contain errors of recognition.

Barbara.
"PS -- this note was dictated entirely in DNS, and without any correction on my part, they're free can see several errors that have occurred.

I will agree that DNS 9.0 Preferred seems to have its good days and bad days. However, I have only been using it for a few weeks and I figured it would get better as we got used to each other.

Answering this post got me looking for tips on improving speech recognition. I found this article which lists tips for improving speech recognition. There was some food for thought here:

http://www.emicrophones.com/docDetails.asp?DocumentID=38

1. I hadn't given much thought to my sound card. I ought to look at that.
2. I haven't tried a better microphone (I read reviews on Amazon that said to throw away the one that comes with the program, but I have kept using it).
3. There were some microphone positioning tips I hadn't tried.
4. I haven't looked at any forums related to improving accuracy, and there are several, believe it or not. I actually found several helpful tips related to some lagging in the program when you first start up. Here's one:

http://knowbrainer.com/PubForum/index.cfm?page=viewForumTopic&topicId=2412

5. I haven't trained much with the included scripts, because "everyone" says it's not necessary now. However, the training I did might have been done with a crummy mic or sound card, and might not have been very helpful also?

Option I *might* examine, if some of these tips don't work, and I feel like parting with a thousand dollars some day.
1. Springing for the full Medical version (although you can't upgrade Preferred to Medical, and I'd just have to eat the $200 for the Preferred Version).
2. And/or add Trigram vocabulary:
http://www.dragon-medical-transcription.com/trigram_medical_vocabulary.html

This was a good question. Made me think about ways to improve my situation, as well.

Brian Cotner, M.D.

Two more easy tips I noticed:

1. I am going to try running Audio Setup daily for a while.
2. I am going to try listening to myself with the playback function and see what I sound like through my present mic and soundcard; that would seem like a "quick and dirty" evaluation of my setup. I don't know why I didn't think to try that myself!

Brian Cotner, MD

I have been using Dragon 8&9 for about a year. I picked them up on ebay. I am using labtech headphones. Both are preferred editions.

I am very happy with the level of transcription. Usually not more than one word. More than a couple words and it would be easier to type.

However, most of my visits are template driven or only need minimal additions. It is not worth the time to put the headphone on, turn on the microphone talk, shift boxes, add templates, possibly turn off microphone. It's easier to type. I probably only use it once a day.

I do understand that in Pediatrics a lot more is template driven than in some other specialities.

I do keep dragon running on one computer all the time and do not seem to have memory issues. Both computers have 1 G of memory.