Hello, my decoding after training the model gives me %15 WER but when I tried with another test set (which was randomly splitted with the decoding dataset) on my own with this command:
pocketsphinx_batch \
-adcin yes \
-cepdir wav \
-cepext .wav \
-ctl test.fileids \
-lm <your.lm> \ # for example en-us.lm.bin from pocketsphinx
-dict <your.dic> \ # for example cmudict-en-us.dict from pocketsphinx
-hmm <your_hmm> \ # for example en-us
-hyp test.hyp
word_align.pl test.transcription test.hyp
I get %45 Error. I think I have a quıite efficient model but I can't use it.
If I am giving wrong parameters for pocketsphinx_batch, what parameters should I give, where can I look for it? There are a lot of logs for decoding, I copied one's parameters that is attached to test but that also gave around %45 error.
Some part of an example decoding Log file is attached.
Hello, my decoding after training the model gives me %15 WER but when I tried with another test set (which was randomly splitted with the decoding dataset) on my own with this command:
pocketsphinx_batch \
-adcin yes \
-cepdir wav \
-cepext .wav \
-ctl test.fileids \
-lm
<your.lm>\ # for example en-us.lm.bin from pocketsphinx-dict
<your.dic>\ # for example cmudict-en-us.dict from pocketsphinx-hmm
<your_hmm>\ # for example en-us-hyp test.hyp
word_align.pl test.transcription test.hyp
I get %45 Error. I think I have a quıite efficient model but I can't use it.
If I am giving wrong parameters for pocketsphinx_batch, what parameters should I give, where can I look for it? There are a lot of logs for decoding, I copied one's parameters that is attached to test but that also gave around %45 error.
Some part of an example decoding Log file is attached.
Thanks.