KnowBrainer Speech Recognition
Decrease font size
Increase font size
Topic Title: one vs 1
Topic Summary: wrong every time
Created On: 09/14/2022 10:34 AM
Status: Post and Reply
Linear : Threading : Single : Branch
 one vs 1   - dcneuro - 09/14/2022 10:34 AM  
 one vs 1   - Alan Cantor - 09/14/2022 10:59 AM  
 one vs 1   - dcneuro - 09/14/2022 11:19 AM  
 one vs 1   - dilligence - 09/14/2022 12:24 PM  
 one vs 1   - gasolo - 09/28/2022 10:57 AM  
 one vs 1   - dcneuro - 09/14/2022 02:57 PM  
 one vs 1   - Mav - 09/15/2022 02:07 AM  
 one vs 1   - Ag - 09/16/2022 08:05 PM  
 one vs 1   - Mav - 09/19/2022 02:46 AM  
 one vs 1   - mikedexter - 09/27/2022 07:11 PM  
 one vs 1   - Alan Cantor - 09/27/2022 08:41 PM  
 one vs 1   - Ag - 03/27/2023 07:59 PM  
Keyword
 09/14/2022 10:34 AM
User is offline View Users Profile Print this message

Author Icon
dcneuro
Junior Member

Posts: 24
Joined: 10/02/2012

I'm not sure if others find this as well or how many even dictate numbers as I do, but whenever I dictate weights or blood pressures, rather than typing "165 pounds" or "125/73", rather, I get "one 65 pounds" or "one 25/73". Also 100% comes out as one hundered %. No matter how many times I correct this it does the same thing. I've also tried adding the digits 0-9 and training them. I've run specific document training with many prior reports with similar phraseology many times, no help.

Thanks in advance.



-------------------------
Dr. S.
 09/14/2022 10:59 AM
User is offline View Users Profile Print this message


Alan Cantor
Top-Tier Member

Posts: 4615
Joined: 12/08/2007

Check your AutoFormatting options. Number formats may have been accidentally changed.

I recently learned (on this forum) that it might be best to strip out numbers from the writing samples given to Dragon to analyze.

Adding 0 through 9 probably won't help. It may have the opposite effect.

If the problem persists, consider creating a new profile. The entire process should only take five or ten minutes.
 09/14/2022 11:19 AM
User is offline View Users Profile Print this message

Author Icon
dcneuro
Junior Member

Posts: 24
Joined: 10/02/2012

Auto formatting options remain unchanged. Actually adding numbers 0-9 has helped considerably as it gets it correct maybe 25% of the time compared to before it didn't get any right. This is a new profile. Like the third time I tried. Does the same thing with each new profile.

-------------------------
Dr. S.
 09/14/2022 12:24 PM
User is offline View Users Profile Print this message

Author Icon
dilligence
Top-Tier Member

Posts: 1683
Joined: 08/16/2010

Taking "165" as an example I'm getting it right each time by saying "one hundred and sixty five"

 

If that does not work for you then you should probably make it a habit of prefixing numbers with "numeral" i.e "numeral 165". 

 

I found that after prefixing with "numeral" it doesn't really matter if I say "one six five", "one sixty five" or "one hundred and sixty five", Dragon® then gets it right every time.

 

Maybe worth a try.   



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

https://speechproductivity.eu


Turbocharge your Dragon productivity with 40+ Power Addons


Speech Productivity Forum

 09/28/2022 10:57 AM
User is offline View Users Profile Print this message


gasolo
Senior Member

Posts: 139
Joined: 01/05/2008

Rob,

That's exactly what I have been doing for years. If I need numbers, I always say"numeral" followed by the number.

 



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

GaSolo

 09/14/2022 02:57 PM
User is offline View Users Profile Print this message

Author Icon
dcneuro
Junior Member

Posts: 24
Joined: 10/02/2012

Good tip. I will try that.

Thanks.

-------------------------
Dr. S.
 09/15/2022 02:07 AM
User is offline View Users Profile Print this message

Author Icon
Mav
Top-Tier Member

Posts: 723
Joined: 10/02/2008

Controlling how your numbers are written (as words or digits) has nothing to do with recognition itself, since you always say the exact same words.

Correcting won't help as well, since Dragon _did_ recognize "one hundred" when you actually said "one hundred".

Formatting to your preference is done by something called "Inverse Text Normalization" (ITN) and the rules for this can be configured under "Auto formatting".

Set "Numbers, if greater than or equal to" to 0 and you should get your numbers written as digits most of the time.

 

In addition, you can vary the written form of words in combination with numbers by looking at the vocabulary editor.

For example, the word "percent" is written as "percent", except if it's being used after numbers, where "%" is written instead. You can even control if numbers are written as words or digits in combination with other words, so for example you can set the word "pound"'s properties to write preceding numbers as digits.

If you do this and remove the check from "Numbers, if greater...", you'd get "one hundred sixty-five", but "165 pounds".

Funny enough, with my DLG 15.71 I get "hundred percent" for saying "hundred percent", but "100%" (note the missing blank as well) for saying "onehundred percent".

So obviously, Dragon is missing "hundred" alone from the grammar used to recognize numbers.

Unfortunately, you cannot fix this by simply changing the written form of "hundred" to "100". If you do, Dragon will still write "100 percent", proving that number recognition is being performed before applying the written form.

 

hth

mav

 09/16/2022 08:05 PM
User is offline View Users Profile Print this message

Author Icon
Ag
Top-Tier Member

Posts: 1150
Joined: 07/08/2019

What they said, but sympathizing:

Like @dcneuro, I often say "one sixty five" expecting "165", but was disappointed to get "one 65".

Saying "one hundred and sixty five" reliably gave "165". but it takes time to say "hundred", etc.

eventually I gave up and just set Dragon Autoformatting options to always emit numbers if >0 => i.e. always. so now I reliably get "one sixty five"-->"165".

as a result of this I always have to use an alternate spoken form in order to get a number spelled out, but fortunately "one as word"-->"one" works well enough.

unfortunately try as I might I always get 1st 2nd 3rd. Spoken forms do not seem to help: "first\\first as word" still always produces 1st. Similarly for 2nd. that is annoying when I talk about time units.

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

DPG15.6 (also DPI 15.3) + KB, Sennheiser MB Pro 1 UC ML, BTD 800 dongle, Windows 10 Pro, MS Surface Book 3, Intel Core i7-1065G7 CPU @ 1.3/1.5GHz (4 cores, 8 logical, GPU=NVIDIA Quadro RTX 3000 with Max-Q Design.

 09/19/2022 02:46 AM
User is offline View Users Profile Print this message

Author Icon
Mav
Top-Tier Member

Posts: 723
Joined: 10/02/2008

Originally posted by: Ag [...] unfortunately try as I might I always get 1st 2nd 3rd. Spoken forms do not seem to help: "first\\first as word" still always produces 1st. Similarly for 2nd. that is annoying when I talk about time units.

Dragon seems to stop processing words due to their properties once automatic number formatting has been applied.

If you set "Numbers as digits..." to the default (2 for English, 13 for German), properties you set for ordinal words are ignored (e.g. changing the written form of "second" to "2." and having following words to be capitalized).

I'd always get "2nd" when there is no word with this written form shown in the vocabulary.

When I set the value to 10, I correctly get "2." with following capitalization.

Dragon's hidden number formatting rules seem to be a bit inconsistent with the rest.

 

hth

mav

 09/27/2022 07:11 PM
User is offline View Users Profile Print this message

Author Icon
mikedexter
New Member

Posts: 2
Joined: 02/08/2022

I have a colleague who is also experiencing this problem, but I don't think it's an auto-formatting issue (we've triple checked the settings). It seems to be more like Dragon is confusing the pronoun 'one' with the number 'one'. For example, instead of "no one" they get "no 1". Instead of "$1.80" they get "one dollar $0.80".

Is correction likely to help with this? Or is it a corrupt profile? The person's profile was only made about a month ago.
 09/27/2022 08:41 PM
User is offline View Users Profile Print this message


Alan Cantor
Top-Tier Member

Posts: 4615
Joined: 12/08/2007

Before creating a new profile...

1. Instead of saying "no one" in isolation, give Dragon more context by saying the words as part of a longer utterance, e.g., "No one I know speaks in two hyphen word phrases exclamation mark".

2. Try "one dollar AND eighty cents" instead of "one dollar eighty cents."
 03/27/2023 07:59 PM
User is offline View Users Profile Print this message

Author Icon
Ag
Top-Tier Member

Posts: 1150
Joined: 07/08/2019

I finally got annoyed enough to do the following:

For a long time I have had my Dragon autoformatting options set to

ON: numbers, if greater than or equal to 0
ON: Other numbers

(also ON: Allow pauses in formatted phrases, which may or may not matter, the another post about the 123 123456 issue/interaction between DragonCapture and standard Dragon autoformatting)

reason: I tend to be number oriented, so I'm happier living with "1" and "1st" than sporadic number words and long number strings "11221"

but occasionally I really do want the word forms of numbers

hence I just wrote a command that takes things like

"one like word" --and-gives-me--> one
"fifth like Number word" --and-gives-me--> fifth

---

Let's see how happy this makes me....

---

so far, relatively happy. biggest annoyance is as usual with using commands to build pseudo-vocabulary rather than actual vocabulary/custom words: bigger delays, Longer pauses required to separate the command from other dictation.

The usual issue with blank insertion - do I want the command output have blanks around it or not. since this is verbose wordy stuff, I am erring on providing excessive space before and after. Excessive space reduces readability less than not having space. (Not actually when you are dictating digits)

===

BTW: prior to this, I have tried repeatedly to create custom words/vocabulary forms such as

first/first like word
second/second like word

but these failed, because apparently autoformatting is applied after the custom word recognition.

it turns out that I had left the "1st/2nd like word" entries in my custom words.

As a result my new commands did Not work for " first/second like word" spoken forms. The custom word/vocabulary took precedence. Worst, when I deleted these older custom word forms, I started getting the indication of a problem.

I think this clearly answers something I have long wondered about: when you delete something that has been in your vocabulary for quite some time, with fairly high probability, Dragon does not necessarily fix things up. Sometimes it will leave error nodes around. I have noticed this before, and have seen that the problem usually goes away the next time I build a profile from scratch without the custom words that were deleted. But this is a fairly good example, although I expect people to tell me I'm wrong.

Fortunately, I have alternate forms that work "first/second like number word".

===




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

DPG15.6 (also DPI 15.3) + KB, Sennheiser MB Pro 1 UC ML, BTD 800 dongle, Windows 10 Pro, MS Surface Book 3, Intel Core i7-1065G7 CPU @ 1.3/1.5GHz (4 cores, 8 logical, GPU=NVIDIA Quadro RTX 3000 with Max-Q Design.

KnowBrainer Speech Recognition » Dragon Speech Recognition » one vs 1

Statistics
32616 users are registered to the KnowBrainer Speech Recognition forum.
There are currently 0 users logged in.
The most users ever online was 12124 on 09/09/2020 at 04:59 AM.
There are currently 265 guests browsing this forum, which makes a total of 265 users using this forum.

FuseTalk Standard Edition v4.0 - © 1999-2023 FuseTalk™ Inc. All rights reserved.