Hi, I am working with pocketsphinx using keyword search.
I am trying to make something like a voice-controlled contacts app for android. But to improve the pocketsphinx accurace I need to get the full list of scores instead of just the best one. For example, if I say "Maria", and in my contacts list has Maria,Mary,May,Mara I want to get a list of scores, one for each name, and this way I can post-process that probabilities list with my own context-dependent algoritms, for example, if It is friday and I usually talk fridays with Mara the "weight" of the context probability will increase the overall score of that contact. etc.
I have tried decoder.nbest() and decoder.seg() with no luck. With the first method I get allways nulll, and with seg() I get at most only 2 scores, This is my feat.params file:
-lowerf 130
-upperf 6800
-nfilt 25
-transform dct
-lifter 22
-feat 1s_c_d_dd
-svspec 0-12/13-25/26-38
-agc none
-cmn current
-varnorm no
-model ptm
-input_endian big
I am using only the keyword search, using a file with a contact per line, and a default weight
Any help will be appreciate. Thanks
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi, I am working with pocketsphinx using keyword search.
I am trying to make something like a voice-controlled contacts app for android. But to improve the pocketsphinx accurace I need to get the full list of scores instead of just the best one. For example, if I say "Maria", and in my contacts list has Maria,Mary,May,Mara I want to get a list of scores, one for each name, and this way I can post-process that probabilities list with my own context-dependent algoritms, for example, if It is friday and I usually talk fridays with Mara the "weight" of the context probability will increase the overall score of that contact. etc.
I have tried decoder.nbest() and decoder.seg() with no luck. With the first method I get allways nulll, and with seg() I get at most only 2 scores, This is my feat.params file:
-lowerf 130
-upperf 6800
-nfilt 25
-transform dct
-lifter 22
-feat 1s_c_d_dd
-svspec 0-12/13-25/26-38
-agc none
-cmn current
-varnorm no
-model ptm
-input_endian big
I am using only the keyword search, using a file with a contact per line, and a default weight
Any help will be appreciate. Thanks