Re: [simon-development] Acoustic Model Switching

[Peter G. <gr...@si...>]

Hi Adam,

On Tuesday 29 November 2011 20:50:26 Adam Nash wrote:
> I finished the stuff that lets the user specify the conditions that will
> trigger certain sample groups/acoustic models.  I put it in the
> recording sub-section of the simon config menu - let me know if you want
> it elsewhere.  Also, let me know of any usability tweaks that you think
> should be done.

The recording settings are getting a bit crowded but I can't think of any 
place more suitable either.
Usability: IMHO it would be nice if one could enter the sample group in the 
add-condition dialog. Also - and I think this is really important - the sample 
group selection should be a drop-down filled with:
a.) The sample groups of the current sound configuration
b.) The sample groups of the prompts
c.) The previously configured sample group

But the context thing is clearly only going to be set up by power users 
anyway...

> I got and error message about the monophones being unavailable when the
> language model context was changing.  According to this debug output I
> guess that the file was there, it was just being used elsewhere.  I
> don't think that this happened before:
I can't really test it now as I don't have a valid model here but it was a 
trivial bug -> Likely fixed.

> Also I'm getting this error even though I have training samples for all
> of my vocabulary:
> 
> "Could not start recognition because the system reports that the
> recognition is not ready.
> Please check if you have defined a vocabulary, an appropriate grammar
> and recorded a few trainings samples.
> The system will then, upon synchronization, generate the model which
> will be used for the recognition."
Well this is to be expected if your model generation fails. Basically this is 
the error for "no active model yet".

> I have it so that the sample groups will be sent to the server now when
> they change, but I have not been able to test it because of these
> errors.  They may be due to something that got mucked on my system, so I
> will try to do some troubleshooting.
Nope, sorry those were upstream fixes that I've probably introduced when I fixed 
a couple of other issues with models that only contain single words for our 
upcoming Italian speech model.

Best regards,
Peter