[TriLUG] FOSS voice assistant framework

[Phillip Rhodes (Fogbeam Labs) via TriLUG]

I have not yet, but I had a similar idea recently... get voice reco running
on an rPi and build sort of a DIY alexa.  It's something I want to explore,
but just haven't had a lot of time to put into it.   But FWIW, my earliest
thought was to use CMU Sphinx[1] for the voice recognition part.

If anybody starts on a project like this, please do share with the list
from time to time... I know I'd love to hear about whatever results you
achieve. And I imagine others will as well.


[1]: https://cmusphinx.github.io/

On Tue, Jul 18, 2017 at 10:32 AM, Jack Hill via TriLUG <trilug at trilug.org>
wrote:

> On Tue, 18 Jul 2017, Tim Jowers via TriLUG wrote:
>
> Hi,
>>
>> Has anyone worked with a FOSS voice assistance framework?  I am not really
>> happy with Siri, Samsung, or Alexa from the developers perspective. I
>> imagine one could setup something with a raspberry pi DIY etc so people
>> could input voice and get back help.
>>
>> To me, seems like a great opportunity for collaboration and also for FOSS
>> to provide revenue streams to developers. One issue with the big companies
>> is they tend to take over any good ideas and bundle them. ("bundling"
>> technique of market dominance)
>>
>> Also not happy from the user's perspective as Alexa has exposed my 6 year
>> old and several neighbors repeatedly to curse words and no telling what
>> concepts. I called Amazon and they said they would work on it if they get
>> enough requests. At this point, had to disallow usage unless parents are
>> present. Samsung voice assistance is really weak, of course.
>>
>> Been years since we touched on this stuff in college; so, let me know if
>> there is any formal collaboration on language modeling/ontology stuff,
>> too.
>>
>> Thanks in advance,
>> Tim
>>
>
> Tim,
>
> Just yesterday, I became aware of two projects to create free machine
> learning models of voice recognition:
>
> https://voice.mozilla.org/
> http://www.voxforge.org/
>
> As I understand it, the idea is that speech recognition software needs a
> huge corpus of training data to work well (which is exactly what Google and
> Amazon have). With some extra constraints (like a limited vocabulary) you
> may be able to get by with less training data.
>
> As far as actual utilities that make use of voice recognition, I was able
> to find
>
> https://cmusphinx.github.io/2015/01/ila-voice-assistant/
> https://cmusphinx.github.io/2014/04/jasper-personal-assistan
> t-for-raspberry-pi/
>
> I haven't use them, so can't make a recommendation, but it could be
> interesting.
>
> Best,
> Jack
>
> --
> This message was sent to: Phillip Rhodes <fogbeam at gmail.com>
> To unsubscribe, send a blank message to trilug-leave at trilug.org from that
> address.
> TriLUG mailing list : https://www.trilug.org/mailman/listinfo/trilug
> Unsubscribe or edit options on the web  : https://www.trilug.org/mailman
> /options/trilug/fogbeam%40gmail.com
> Welcome to TriLUG: http://trilug.org/welcome
>