Recogntion problems

[toto313]

I have a problem with a certain order
here I use the beautiful plugin for MSFS of jamescl604 on the command "com un swap" VA understands it easily on other example "altitude sur 500" on the LOG written unrecognized:'altitude sur 500'

images to better explain



com un swap



altitude sur 500

[Pfeil]

What language is the speech recognition engine you're using designed for?

[toto313]

Hi Pfeil

French language

PS: if. When i Say: altitude sur :    it's reconized

[Gary]

If you are saying, 'altitude sur 500', there is no command that corresponds to that, according to the screenshots you have provided:

[Autopilot;Auto pilot] altitude [sur;au;at;to;and] *

You should be able to say, 'autopilot altitude sur 500' given that dynamic phrase.

You can try changing the phrase to one of the following:

[Autopilot;Auto pilot;altitude][sur;au;at;to;and] *

Or

altitude [sur;au;at;to;and] *



Of course, that may be total nonsense since I'm not entirely sure what the context means (combined with not knowing French).   However, I hope that steers you in the right direction!

[toto313]

the problem was very simple.
i just have to train the voice to voice recognition several times a "tip maybe" for novice like me in voiceattack go to setting --> recognition -->utilities--> add remove dictionary word--> and add word
For example the word "altitude" is not recognized by voice attack, you have to add the word "altitude" in the dictionary and repeat the training several times. look here https://forum.voiceattack.com/smf/index.php?PHPSESSID=2946257311ca8831a2de6c22aef496ae&topic=3474.msg16557#msg16557

[Gumbie]

So I am having a similar issue but I am really having issues understanding why but when I use the word Kneeboard  Speech Recognition understands the word 90% but voice attack thinks I am saying Need Board, what the heck is this???   
  If speech recognition is understanding the word almost 100% why is voice attack completely filling in the blank with non sense like Need Board.     I get sounds similar but if Recognition is having very little issue understanding the word at 90% + correctly why is voice attack not and blowing a massive rhino to do the same.    I really wish it was something as basic as training but its not I have trained all the words used an still having it mistake words in voice attack,  but speech recognition still getting above 90% for the same few words.      Any advise besides using different keyword for commands because I have tried this and seems that still getting same issue. Example     Board was used instead of Kneeboard and Voice attack reads Board as Boar at 86% unrecognized but speech gets in 90% + correct again which I find funny because doesn't the program rely on speech recognition.

[Pfeil]

Do you have a command in your profile with the phrase "kneeboard"? If so, what are the contents of the "When I say" field for that command?


By "Speech Recognition", are you referring to the Windows Speech Recognition application?

[Gary]

@Gumbie - If you are using Windows' speech app as Pfeil has indicated, the speech engine that you are using is exactly the same, with the same training.  The only difference is that the instance of the speech engine that VA uses is focused based on the commands that you provide.  For example, if you provide a command called, 'kneeboard', VA cranks up an instance of the speech engine and tells it to focus itself around, 'kneeboard'.  If you say, 'kneeboard' and it comes back as, 'need board', that means that the speech engine, although being told to focus on 'kneeboard' is having trouble with this situation for some reason.  There could be any number of factors contributing to that situation that you can read about in this thread:  https://forum.voiceattack.com/smf/index.php?topic=1635.0    (It's a bit of a read, but it has helped many over the years both understand what is going on, but also help prepare or fix their PCs to work better with Windows' speech recognition).  Again, it's a very simple mechanism - give the speech engine a set of words to focus on.  Outside of that, the speech engine is rather blind to anything else - it kind of makes best guesses to anything else that is said.  So, if it's having trouble negotiating, 'kneeboard' even though it has a direct instruction to do so, it's not getting it for some reason, so, it yields, 'need board' as its best guess.  It could be anything from hardware to command weight to speaking accent to installed drivers and software.  You'll want to take a run through that thread to see if anything might help.  I did notice that you've had VA for a rather long time - if this is a new situation, you'll really want to look at hardware, drivers and noise reduction software as those can trip you up unexpectedly.

Given all that, and as a testimony to environment being a factor, I added a command just now called, 'kneeboard'.  It hits at 100% (with 20 tries) with a confidence of 95 and above without any additional training.  I've had cases where the speech engine had real trouble with one of my commands - no matter what I did, it couldn't seem to get the word right even with additional training.  What I did to solve the problem was look at what the speech engine was favoring as a result and using it, as it doesn't cost much to do this.  In your case, I would make a command where the spoken phrase looks like 'kneeboard;need board;need boar' - that will trigger the command for any of those.  I know that's not what you want to hear, but again, there's a reason why it's not getting it and it's going to boil down to something in your setup.  This is a way to get around that if you don't want to spend a bunch of time trying to figure it out.

Hope that helps!