I have lately been wishing for more accuracy with UK English when using the
default acoustic model. This wiki page states that this is a case for adapting
the default acoustic model (especially since I don't have the time or audio
available to train):
"Adaptation is known to work well when you are using different recording
environments (close-distance or far microphone or telephone channel), or when
a slightly different accent is present (UK English or even Indian English) or
even another language."
When I follow this up and go to the adaptation tutorial page, I don't see
information on how to adapt the default acoustic model from US English to
another accent, just information on how to adapt the model to an individual's
speech:
Am I overlooking the information, or is it at another URL, or is my
understanding of the problem incorrect and there isn't a way to easily adapt
the default English acoustic model to UK English without doing adaptation
based on audio from individual speakers?
As always, thanks very much for your assistance,
Halle
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Yes, I saw those models but I don't think they are an option for me, since the
maintainer doesn't seem to want them used commercially and the original
database used to create them is for research use only. Do you happen to know
of a representative UK English adaptation set that isn't exclusively for use
in research?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Nickolay,
I have lately been wishing for more accuracy with UK English when using the
default acoustic model. This wiki page states that this is a case for adapting
the default acoustic model (especially since I don't have the time or audio
available to train):
http://cmusphinx.sourceforge.net/wiki/tutorialam
"Adaptation is known to work well when you are using different recording
environments (close-distance or far microphone or telephone channel), or when
a slightly different accent is present (UK English or even Indian English) or
even another language."
When I follow this up and go to the adaptation tutorial page, I don't see
information on how to adapt the default acoustic model from US English to
another accent, just information on how to adapt the model to an individual's
speech:
http://cmusphinx.sourceforge.net/wiki/tutorialadapt
Am I overlooking the information, or is it at another URL, or is my
understanding of the problem incorrect and there isn't a way to easily adapt
the default English acoustic model to UK English without doing adaptation
based on audio from individual speakers?
As always, thanks very much for your assistance,
Halle
Hello
Adaptation for accent is not different from adaptation to the speaker. Just
adaptation set needs to be more representative.
There are UK English models here:
http://www.keithv.com/software/sphinx/uk/
Yes, I saw those models but I don't think they are an option for me, since the
maintainer doesn't seem to want them used commercially and the original
database used to create them is for research use only. Do you happen to know
of a representative UK English adaptation set that isn't exclusively for use
in research?
Unfortunately no
All righty, thanks for your input in any case!