I'm using pocketsphinx in a project I'm working on with some good success.
However, I've seen lots of references to getting an "accuracy" rating
returned. Does pocketsphinx return some accuracy percentage or is everyone
just manually calculating that? I've looked in the documentation and data that
pocketsphinx spits out after transcribing something, but have not seen
anything that looks like accuracy. I'd like to use something like this to
programatically compare 2 transcription methods
Is there a way I can get pocketsphinx to guess at how confident it was in a
transcription or is there some returned value I'm missing that does something
similar?
Thanks!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'd like to use something like this to programatically compare 2
transcription methods Is there a way I can get pocketsphinx to guess at how
confident it was in a transcription or is there some returned value I'm
missing that does something similar?
In statistics those two tasks (compare two and get confidence) are completely
different and need to be addressed differently. First could be archived with
the comparision of the score of the path (value returned by ps_get_hyp
function). Second require recognizer to have a large amount of competing
hypothesis in order to get meaningful number (value returned by ps_get_prob
function). For details on the functions mentioned you can check API
documentation.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
N00b question:
I'm using pocketsphinx in a project I'm working on with some good success.
However, I've seen lots of references to getting an "accuracy" rating
returned. Does pocketsphinx return some accuracy percentage or is everyone
just manually calculating that? I've looked in the documentation and data that
pocketsphinx spits out after transcribing something, but have not seen
anything that looks like accuracy. I'd like to use something like this to
programatically compare 2 transcription methods
Is there a way I can get pocketsphinx to guess at how confident it was in a
transcription or is there some returned value I'm missing that does something
similar?
Thanks!
In statistics those two tasks (compare two and get confidence) are completely
different and need to be addressed differently. First could be archived with
the comparision of the score of the path (value returned by ps_get_hyp
function). Second require recognizer to have a large amount of competing
hypothesis in order to get meaningful number (value returned by ps_get_prob
function). For details on the functions mentioned you can check API
documentation.