I am using Sphinx 3 for decoding.
I am using Fedora for creating the language model. When I run the following
command: perl scripts_pl/decode/slave.pl, I get the following message:
word_align.pl failed with error code 65280 at scripts_pl/decode/slave.pl line
172.
The log file shows the following errors:
ERROR: "cont_mgau.c", line 666: Weight normalization failed for 12 senones
ERROR: "srch_time_switch_tree.c", line 849: ERROR Fr 3, best HMM score > 0 (2013284884); int32 wraparound?
ERROR: "wid.c", line 282: UNK is not a word in dictionary and it is not a class tag.
What could be causing these errors?
List item
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
None of this errors cause problem you described. Read log more carefully to
get the reason and learn to provide full log when you report about
problems.
As for UNK, you need to use -vocab_type 1 of ingram2lm option to create
closed-vocabulary language model
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am using Windows XP platform to build a multi-speaker model (12 speakers
presently). The speakers were recorded in a normal office environment with
some noise in the background. The speech duration for each speaker is
approximately 15 minutes. I have completed the training successfully.
However I am having a problem while decoding (Sphinx 3 decoder) and I get the
following error message:
wordalign.pl failed with error code 65280 at scripts_pl/decode/slave.pl line
172.
We have tried looking for errors in the grammar file and have also confirmed
that our transcriptions and fileids are aligned. Please let us know what else
we should recheck and what else could be causing the problem.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
but it is still not working. I have tried these values for my previously
trained models and its working. I have also used the following value sets for
my previously trained models and the are working:
The fact you are using changed values makes me think that you changed
something incorrectly. I could only suggest you to start from defaults once
again and apply your changes one by one in order to find problematic one. In
general, without access to your files it's hard to reproduce your problem,
thus it's impossible to help.
If you want to share your files you can always upload them to a public sharing
resource and give us a link.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am using Sphinx 3 for decoding.
I am using Fedora for creating the language model. When I run the following
command: perl scripts_pl/decode/slave.pl, I get the following message:
word_align.pl failed with error code 65280 at scripts_pl/decode/slave.pl line
172.
The log file shows the following errors:
What could be causing these errors?
List item
None of this errors cause problem you described. Read log more carefully to
get the reason and learn to provide full log when you report about
problems.
As for UNK, you need to use -vocab_type 1 of ingram2lm option to create
closed-vocabulary language model
I am using Windows XP platform to build a multi-speaker model (12 speakers
presently). The speakers were recorded in a normal office environment with
some noise in the background. The speech duration for each speaker is
approximately 15 minutes. I have completed the training successfully.
However I am having a problem while decoding (Sphinx 3 decoder) and I get the
following error message:
wordalign.pl failed with error code 65280 at scripts_pl/decode/slave.pl line
172.
I have uploaded my log file at :
http://www.box.net/shared/vdhm3op8cj
My transcription and configuration files are available at:
http://www.box.net/shared/gumpenamg4
We have tried looking for errors in the grammar file and have also confirmed
that our transcriptions and fileids are aligned. Please let us know what else
we should recheck and what else could be causing the problem.
Don't you think that the beams you changed are not correct?
$DEC_CFG_BEAMWIDTH = "1e-500";
$DEC_CFG_WORDBEAM = "1e-080";
Why did you change them?
Thank you for your reply. I have tried using the following values:
$DEC_CFG_LANGUAGEWEIGHT = "21";
$DEC_CFG_BEAMWIDTH = "1e-800";
$DEC_CFG_WORDBEAM = "1e-080";
but it is still not working. I have tried these values for my previously
trained models and its working. I have also used the following value sets for
my previously trained models and the are working:
$DEC_CFG_BEAMWIDTH = "1e-100";
$DEC_CFG_WORDBEAM = "1e-080";
$DEC_CFG_BEAMWIDTH = "1e-500";
$DEC_CFG_WORDBEAM = "1e-080";
$DEC_CFG_BEAMWIDTH = "1e-900";
$DEC_CFG_WORDBEAM = "1e-080";
Can you suggest some values I should be using? or point out any mistakes that
I may be making in this process?
Defaults are
$DEC_CFG_BEAMWIDTH = "1e-80";
$DEC_CFG_WORDBEAM = "1e-40";
The fact you are using changed values makes me think that you changed
something incorrectly. I could only suggest you to start from defaults once
again and apply your changes one by one in order to find problematic one. In
general, without access to your files it's hard to reproduce your problem,
thus it's impossible to help.
If you want to share your files you can always upload them to a public sharing
resource and give us a link.
Thank you for your help. It is working on default values.