wintermute-devel team mailing list archive

[Jacky Alcine <jacky.alcine@xxxxxxxxxx>]

So did you record the audio to disc first manually or programmatically
then pass it to the app/API?

What did you say to it? We might need to implement a means of
determining accuracy. Like if I were to say

 "How much wood could a woodchuck chuck if a wood chuck could chuck wood? "

and it'd hear..

"How much would could a wood shunt chuck if a wood duck could duck wood?"

There should be a function (calculus time! :D) for determining the
accuracy. I'm not sure what would it involve (perhaps the number of
correctly matched phonemes in comparison to the actual number of used
phonemes). Such a formula would have to range from 0 to 100 (of
course) and would be required to determine the efficiency of a model.
However, models are always restricted by a domain. I'm going to put up
some docs regarding this soon.

After the launching of the SpeechControl project, however, I feel that
this code could be more efficiently used that project and have it just
call upon SpeechControl for synthesizing text. The only issue with
that is that it adds a dependency (in order to use Wintermute's
default speech recognizer, you have to have SpeechControl installed).
Of course, SpeechControl could be diced into a library, a front-facing
application and a daemon for D-Bus, and that's something to be
considered.

Depends, now, Islam, where you'd like to go.

1) Stick to Wintermute and develop the Voice API (more specifically,
pocketsphinx-wintermute and wintermute-plugin-voice).

2) Switch gears and focus on speech recognition only in SpeechControl

3) Do both (1) + (2) while producing another plug-in (I know, stupid)
called speechcontrol-wintermute. If this is done, we'd have to discuss
whether or not it'd make sense to have a plugin that links to the
library (pocketsphinx-wintermute) and have a plugin to SpeechControl
(a wrapper for that library) or just dump the PocketSphinx plugin to
Wintermute.

It's your call and people are free to leap in and join the discussion
(as always :D)

On Sun, Dec 11, 2011 at 6:57 PM, Islam F. Bahnasy <ifarouk@xxxxxxxxx> wrote:
>
> Hi all,
>
> I was able to recognize recorded voice with PocketSphinx as an experiment.
> We need to make that continuous recognition like dictation so that's the coming part and of course integrate all that with Wintermute.
>
> --
> Sincerily,
> Islam Bahnasy
>
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~wintermute-devel
> Post to     : wintermute-devel@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~wintermute-devel
> More help   : https://help.launchpad.net/ListHelp
>
One Vision. One Purpose.
We are the Synthetic Intellect Institute
contact@xxxxxxxxxx :: http://www.thesii.org/ :: #sii on FreeNode

Jacky Alcine, Head Developer of Wintermute
http://jackyalcine.co.cc :: jacky.alcine@xxxxxxxxxx ::
https://launchpad.net/~jackyalcine