Wildcard not working using Portuguese - Brasil Language

[saintchelo]

Hi

I'm using Microsoft Server Speak Recognition Language - TELE (pt - BR) as a Speech Platform Speech Engine and I can't use wildcards - see the screenshot below

In my case, when I speak "something" "sapato" the only thing that is recognized is "sapato", what comes before is not recognized.

How could I use wildcards in my language which is Portuguese-Brazil ?

[Gary]

Hi, Saintchelo

Unfortunately, I'm finding that this is another limitation of the alternate speech engines (you had indicated that you are using Portuguese, so I am assuming that you are using the Portuguese alternate engine).  What is happening is the speech engine is not passing back the full text in that context (that type of speech engine only picks out 'sapato' and that is all).  This is news to me - I will update the documentation to reflect this.   

Sorry about that.  Wildcards are largely unsupported, as they're are only useful in very specific circumstances (which would probably mostly involve things that are not critical).  Thank you for pointing this out.

[saintchelo]

Hi, Saintchelo

Unfortunately, I'm finding that this is another limitation of the alternate speech engines (you had indicated that you are using Portuguese, so I am assuming that you are using the Portuguese alternate engine).  What is happening is the speech engine is not passing back the full text in that context (that type of speech engine only picks out 'sapato' and that is all).  This is news to me - I will update the documentation to reflect this.   

Sorry about that.  Wildcards are largely unsupported, as they're are only useful in very specific circumstances (which would probably mostly involve things that are not critical).  Thank you for pointing this out.

Thank you for your attention

Understood.

Another question please.

Is there any way to store what I say even if it is not a command that is in my profile?

For example: in my profile I only have one command which is "shoe" and when I say "shoe" the actions related to the command "shoe" are executed, but if I say for example "shirt" or any other phrase that is not in my profile, would be a way to store this word "shirt" so I can use it later?

[Pfeil]

Speech Platform 11 (the alternative engine you're using) does not have dictation capabilities, meaning it can effectively only recognize predefined commands, and cannot provide information on input that isn't a command.

Unfortunately, Microsoft does not provide a SAPI engine (which does support dictation) for your language.


So, in short: no.

[saintchelo]

Speech Platform 11 (the alternative engine you're using) does not have dictation capabilities, meaning it can effectively only recognize predefined commands, and cannot provide information on input that isn't a command.

Unfortunately, Microsoft does not provide a SAPI engine (which does support dictation) for your language.


So, in short: no.

I understood. But when I say a word or phrase that is not in my commands, where does that word or phrase go? Because that word or phrase must go somewhere to compare it to my commands

[SemlerPDX]

Speech Platform 11 (the alternative engine you're using) does not have dictation capabilities, meaning it can effectively only recognize predefined commands, and cannot provide information on input that isn't a command.

Unfortunately, Microsoft does not provide a SAPI engine (which does support dictation) for your language.


So, in short: no.

I understood. But when I say a word or phrase that is not in my commands, where does that word or phrase go? Because that word or phrase must go somewhere to compare it to my commands

You could try checking the value of the token {LASTSPOKEN} -- but if I'm understanding what Pfeil wrote about Speech Platform 11, that may not work for your case.

[Pfeil]

As far as I'm aware, Speech Platform 11 doesn't use words for recognition, but rather phonemes (the sounds that form words, essentially), which it then attempts to match to the phonemes that form the predefined words or phrases in its command list.

Matching to a known set of possibilities is a lot simpler (though still complex, in this context) than matching to any possible word in a language


Perhaps think of like being given a list of a few words in a language you don't speak, like Chinese, for example, and recordings of how they sound.
Then, if you were to listen to a native speaker, you might be able to pick up those few words you know, but anything other than those words might as well be gibberish to you, as you have no idea what the sounds you're hearing signify.


So in practice, for Speech Platform 11, if it doesn't find a match in the predefined list of commands, it discards the input, as it has no capability of processing that input further. I.E. it can only compare against the commands it has been provided with.