Problems with dictating any date in February.   (NB. I use UK/NZ date format of day-month-year).

For dates I say  " eight nine twenty-two" and I get 8/9/22, which is perfect.

However, the month of february doesn't work,  I just get a "-" instead of the 2. 

Eg. 1st of feb 2023,   I say "one two twenty-three" and I get 1-23, instead of 1/2/23.     The "two" gets interpreted as a "-".

In each instance I have to say "select that",  then I get the option of "1/2/23" in the "alternative options box". 

So it does get a bit tiresome.  However much I correct, it never learns.

Any work arounds?   Any help appreciated. 

 

 

when I say "one two twenty-three" on my DMO, it outputs exactly "2023-01-02", which is how I want my dates formatted. I'm using the "American English" version, too.

You may have to play around in the dates and number formatting options in the DMO "Manage Formatting" pane. These can only be accessed through the "Manage Vocabulary..." sub-menu, and not through the regular "Options..." tab.

1 more place you might want to look into is in your own custom vocabularies, to see if you accidentally formatted something where hyphen is written out as "to".

In the end, it may just be your particular instance of DMO's "idiosyncrasies". For example, for the life of me, I could not get DMO to not write out "PRN" as "as needed", whether through explicitly setting DMO not to expand contractions, or by adding "PRN" as a custom phrase - "P R N" (sic). My current workaround is a custom phrase of written form "PRN" for "pro re nata" as spoken form. if anyone has any ideas on how to overcome this one without having me learn pig Latin, please share.

And sometimes, several weeks later, without any user intervention, what formatting shenanigans that plagues you just mysteriously "rights" itself. That was my impression with both DME and DMO.

Desktop Dragon users may deal with the occasional "Gremlin" that rids itself after a simple reboot. DMO users would have to wait for "them" to reboot the servers after some massive solar storm flipped a bunch of bits in Nuance's funny farm.

I use DMPE and have no familiarity with DMO,but IF you do have ability to make Text & Graphic commands in DMO, you could name a commannd "Pea Are An", and for the text output "as needed".

 

MDH

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

Thanks MDH (Mark) for the suggestion. Didn't work, however, as I just tried it now. I recall trying something along the line of "homophone words for letters" before and it didn't work for "PRN" or "NPO" (which DMO/DME insists on outputting "N.p.o.", which is fine by itself, except that as such it doesn't trigger the order string in my EHR ... anyway, comparatively a first-world problem ...)

I am on-call now and just this moment I encountered inexplicable formatting curiosities whereby dictating "heart rate between one forty to one fifty" outputs "... 1 40-1 50". "A hundred forty to a hundred fifty" outputs "100 4250" without fail. Finally "one hundred forty to one hundred fifty" gets me the "140-150" that I am after.

Tomorrow morning after a reboot, somewhere, such bizarro may fix itself. Or maybe it's just me, perpetually struggling to keep up with the pidgin Dragon Nuance speaks.

Try one forty hyphen one fifty which should give you 140-150

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

Change "No" to "Know" w/KnowBrainer 2022
Trial Downloads
Dragon/Sales@KnowBrainer.com 
(615) 884-4558 ex 1

For phrases with two numbers like "140 to 150," I don't think it is reasonable to expect Dragon to parse that because Dragon must decide between "2","to","too", and "two." (Of course, your expectations might be different than mine.)

I learned to say things like "from 140 up to 150" to inject a non-number word between the two numbers. Or "from 140 hyphen 150" or (for dates) "one slash two number slash twenty-three" to get 1/2/23.

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

Win10/11/x64, AMD Ryzen 7 3700X/3950X, 64/128GB RAM, Dragon 15.3, SP 7 Standard, SpeechStart, Office 365, KB 2017, Dragon Capture, Samson Meteor USB Desk Mic, Amazon YUWAKAYI headset, Klim and JUKSTG earbuds with microphones, excellent Sareville Wireless Mono Headset, 3 BenQ 2560x1440 monitors, Microsoft Sculpt Keyboard and Logitech G502 awesome gaming mouse.

^^^ You both make good points, Kevin and Lunis. Certainly the way to guarantee the hyphen is to speak the hyphen.

Now one thing Kevin: I would never speak "140 to 150" as an isolated phrase while expecting Dragon to read my mind as to deciding "142" vs "140 to". That will require the kind of "brainwave AI" Elon is still testing out on his monkeys with.

I always speak in context, as in "Heart rate between 140 to 150". Just that DMO outputs "Heart rate between 1 40-1 50" (sic) for some reason. Its context-aware algorithm or "AI" does decipher correctly that the homophone "to/two/too" here should be "to" and written out as hyphen. Just that the formatting is annoyingly off, for no rhymes or reason.

But I accept that I am bemoaning a first-world problem that can be overcome with a little discipline.

To OP: extending upon the principle of which Lunis suggested, speaking out the "slash" (/) between the digits may ensure the formatting you desire ...

FWIW, even when you say "between 140 to 150..." Dragon must decide between "1","one","40","forty","42","forty-two", and so on.

If it were me, I would say "Heart rate between numeral 140 and numeral 150." Dragon eats the 'numeral' hints and outputs "Heart rate between 140 and 150."

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

Win10/11/x64, AMD Ryzen 7 3700X/3950X, 64/128GB RAM, Dragon 15.3, SP 7 Standard, SpeechStart, Office 365, KB 2017, Dragon Capture, Samson Meteor USB Desk Mic, Amazon YUWAKAYI headset, Klim and JUKSTG earbuds with microphones, excellent Sareville Wireless Mono Headset, 3 BenQ 2560x1440 monitors, Microsoft Sculpt Keyboard and Logitech G502 awesome gaming mouse.

For anyone who is interested, you can say one forty over one fifty to get 140/150

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

Change "No" to "Know" w/KnowBrainer 2022
Trial Downloads
Dragon/Sales@KnowBrainer.com 
(615) 884-4558 ex 1

While what you state is true and good, Kevin, what I am ranting about is simply this: in this day and age when "everyday AI" is driving cars on real streets, steering missiles to real faces, handing in essays to real professors, and spewing out code blocks to real developers, is it too much to expect a company that bills itself as the leading purveyor of "natural language AI" to be able to format a couple of digits correctly when a real (or "fake") doctor spits out a string of numbers immediately following a non-esoteric medical term such as "heart rate"?

Workarounds clearly abound. I also agree with your implicit contention that the key is to lower expectations, exercise some discipline, and accept that there is "AI", and there is AI.

P.S., a colleague was just raving to me today about the amazing ChatGPT that can write out intelligent medical recommendations/assessments on real diagnoses ... and thinking that this ought to be the (near) future of medical record output.

I agree with you in that Dragon should be able to format all those special cases according to the context, but please bear in mind when Dragon was developed.

This was waaaaaay before any so-called AI surfaced.

To my knowledge, they only added neural networks as part of the acoustic model and that's when they started claiming that they use AI.

 

But this definitely is not true for the formatting rules (which are the reason that the recognized phonemes "tu:" are written as "two", "2", "too" or sometimes even "to" or "-").

All of those formatting rules are pretty simple if-then rules which started out as simple string replacements. Now they are taking the context into consideration, but only to a very limited degree (e.g. is a word followed by a number) and on pretty much the same basis of pattern-matching.

 

Given the fact that Nuance has switched to cloud-only mode and are phasing out "classical" Dragon more and more, I strongly doubt that they will invest anything into moving formatting rules to use "real" AI.

 

mav