I would like to ask what are these percentage I see as I progress. During training, I am seeing word error rate which is 27.4%. During decoding using the pocketsphinx_batch and using the test data, I have seen also with a percentage saying the accuracy which is 15.4%. Could someone explain it?
Also in using pocketsphinx_batch or during live recognition, the output words detected are two which my dictionary is using just one word. This is called insertion right? I am building a single digit recognizer only.
Thanks a lot in advance for answering and enlightening me in this particular topic.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I would like to ask what are these percentage I see as I progress. During training, I am seeing word error rate which is 27.4%. During decoding using the pocketsphinx_batch and using the test data, I have seen also with a percentage saying the accuracy which is 15.4%. Could someone explain it?
Also in using pocketsphinx_batch or during live recognition, the output words detected are two which my dictionary is using just one word. This is called insertion right? I am building a single digit recognizer only.
Thanks a lot in advance for answering and enlightening me in this particular topic.
Please see "Word error rate" section in https://cmusphinx.github.io/wiki/tutorialconcepts/