I was getting a small number of hypothesis in the nbest list (sphinx3). So I
increased the beamwidths (i.e. decreased beam, pbeam and wbeam) and now there
are more number of hypothesis.
1. Are there any other param to be fiddled with to get more # hypo. in nbest list ? I was trying to get the kind of nbest scores people report (how many times best hypothesis occurred in the nbest list etc.)
Does the number of nbest hypothesis depend on length of utterance, like more the speech, more # hypo?
(Just asking) Does anybody here have the sourcecode of the method reported in "N-best List Generation using Word and Phoneme Recognition Fusion" by Ernest Pusateri and JM Van Thong
Thanks.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I was getting a small number of hypothesis in the nbest list (sphinx3). So I
increased the beamwidths (i.e. decreased beam, pbeam and wbeam) and now there
are more number of hypothesis.
1. Are there any other param to be fiddled with to get more # hypo. in nbest list ? I was trying to get the kind of nbest scores people report (how many times best hypothesis occurred in the nbest list etc.)
Does the number of nbest hypothesis depend on length of utterance, like more the speech, more # hypo?
(Just asking) Does anybody here have the sourcecode of the method reported in "N-best List Generation using Word and Phoneme Recognition Fusion" by Ernest Pusateri and JM Van Thong
Thanks.