I have followed the Robust Group Tutorial using XP SP2, ActivePerl, VS 2005, Sphinxbase 0.3, pocketSphinx 0.4.1, the nightly SphinxTrain from svn and AN4 database. According with the tutorial I wrote a simple script that is:
However the last script raises the following error
...
INFO: ........\src\programs\wave2feat\wave2feat.c(139): wav/an4test_clstk/mmxg/cen8-mmxg-b.sph
Can't open file C:/Documents and Settings/acorisco/Escritorio/acorisco/DiplomArbeit/an4/result/an4.match
In fact, the result dir has not been created. I also tried using all the nightly version but when compiling PocketSphinx I get several errors (along with hundreds of warnings) about some missing files such as s2_mmap.c
I don't know much about Perl or ASR, I want to learn through the demos and tutorials but that's been very frustrating. Can any one give me a hand on that???
Thanks,
Corisco
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
i am new to sphinx, too. I run into some error when I compiled pocketsphinx.
vc reports there are some files missing duing compiling: s2_mmap.c ,lm.c, cache_lm.c, libpocketsphinx.lib. So, the building result is 0 succesess, 10 failed. Did you see this error before. Any suggestion? thanks alot.
Mandee
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
decode also dumps all errors in log in logdir/decode. Please look there for more precise information. Most probably you need sphinx3_continuous. compile sphinx3 and put executable in bin directory.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Nickolay, thanks for the reply but there's no file in logdir\decode. However, I found that the previous error was due to the fact that I was trying to decode a continuous model with pocketsphinx (semicontinuous decoder). I changed it but now I am getting some errors during training:
MODULE: 20 Training Context Independent models (2007-11-22 12:03)
Phase 3: Forward-Backward
...
This step had 6 ERROR messages and 36 WARNING messages. Please check the log file for details. completed
Current Overall Likelihood Per Frame = -11.7074211020096
Baum welch starting for 1 Gaussian(s), iteration: 2 (1 of 1)
bw Log File
.....
.....
.....
MODULE: 50 Training Context dependent models (2007-11-22 12:44)
Phase 3: Forward-Backward
...
This step had 14 ERROR messages and 0 WARNING messages. Please check the log file for details. completed
Current Overall Likelihood Per Frame = 5.26908373475453
Baum welch starting for 1 Gaussian(s), iteration: 2 (1 of 2)
bw Log File
I have looked through the log files but do not really understand what's happening there. When decoding I get the following error
MODULE: DECODE Decoding using models previously trained
Decoding 130 segments starting at 0 (part 1 of 1)
Could not find executable for %rootdir%\bin\pocketsphinx_batch at %rootdir%\scripts_pl\lib/SphinxTrain/Util.pm line 299.
Can't open %rootdir%/result/an4-1-1.match
SENTENCE ERROR: 100.000% (130/130)
The fact is that neither pocketsphinx.batch nor an4-1-1.match exist (however an4.match does exist).
Whatever help you could bring me will be welcome, thanks!!
Corisco
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Could not find executable for %rootdir%\bin\pocketsphinx_batch at %rootdir%\scripts_pl\lib/SphinxTrain/Util.pm line 299.
The decode process calls decode with test files and language model you are using and decoder produces match file. So you must:
1) compile pocketsphinx_batch
2) put it into %rootdir%/bin
>I have looked through the log files but do not really understand what's happening there.
Well, you can ask about details but not about everything at once. Start with something simple, for example with a single line in the log and ask questions about it.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
However I still get some errors while training in modules
MODULE: 20 Training Context Independent models ........Phase 3: Forward-Backward
MODULE: 50 Training Context dependent ........Phase 3: Forward-Backward
All of them come from gauden.c
ERROR: "........\src\libs\libmodinv\gauden.c", line 1700: var (mgau= 0, feat= 2, density=34, component=0) < 0
ERROR: "........\src\libs\libmodinv\gauden.c", line 1700: var (mgau= 0, feat= 2, density=50, component=1) < 0
ERROR: "........\src\libs\libmodinv\gauden.c", line 1700: var (mgau= 0, feat= 2, density=73, component=2) < 0
I got several errors while compiling the pocketsphinx and sphinxbase nightly versions from svn (due to missed files: cache_lm.c lm.c and s2_mmap.c) so I am using releases 0.4.1 and 0.3 respectively. I am working with XP SP2 and VS 2005
Any clue?? Thanks
Corisco.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi all!
I have followed the Robust Group Tutorial using XP SP2, ActivePerl, VS 2005, Sphinxbase 0.3, pocketSphinx 0.4.1, the nightly SphinxTrain from svn and AN4 database. According with the tutorial I wrote a simple script that is:
chdir("SphinxTrain");
system("perl scripts_pl\setup_tutorial.pl an4");
chdir("..\an4");
system("perl .\scripts_pl\make_feats.pl -ctl etc\an4_train.fileids");
system("perl .\scripts_pl\RunAll.pl");
chdir("..");
chdir("PocketSphinx");
system("perl scripts\setup_tutorial.pl an4");
chdir("..\an4");
system("perl .\scripts_pl\make_feats.pl -ctl etc\an4_test.fileids");
system("perl .\scripts_pl\decode\slave.pl");
However the last script raises the following error
...
INFO: ........\src\programs\wave2feat\wave2feat.c(139): wav/an4test_clstk/mmxg/cen8-mmxg-b.sph
Can't open file C:/Documents and Settings/acorisco/Escritorio/acorisco/DiplomArbeit/an4/result/an4.match
In fact, the result dir has not been created. I also tried using all the nightly version but when compiling PocketSphinx I get several errors (along with hundreds of warnings) about some missing files such as s2_mmap.c
I don't know much about Perl or ASR, I want to learn through the demos and tutorials but that's been very frustrating. Can any one give me a hand on that???
Thanks,
Corisco
Hi Coriscow,
i am new to sphinx, too. I run into some error when I compiled pocketsphinx.
vc reports there are some files missing duing compiling: s2_mmap.c ,lm.c, cache_lm.c, libpocketsphinx.lib. So, the building result is 0 succesess, 10 failed. Did you see this error before. Any suggestion? thanks alot.
Mandee
decode also dumps all errors in log in logdir/decode. Please look there for more precise information. Most probably you need sphinx3_continuous. compile sphinx3 and put executable in bin directory.
Hi again!!
Nickolay, thanks for the reply but there's no file in logdir\decode. However, I found that the previous error was due to the fact that I was trying to decode a continuous model with pocketsphinx (semicontinuous decoder). I changed it but now I am getting some errors during training:
MODULE: 20 Training Context Independent models (2007-11-22 12:03)
Phase 3: Forward-Backward
...
This step had 6 ERROR messages and 36 WARNING messages. Please check the log file for details. completed
Current Overall Likelihood Per Frame = -11.7074211020096
Baum welch starting for 1 Gaussian(s), iteration: 2 (1 of 1)
bw Log File
.....
.....
.....
MODULE: 50 Training Context dependent models (2007-11-22 12:44)
Phase 3: Forward-Backward
...
This step had 14 ERROR messages and 0 WARNING messages. Please check the log file for details. completed
Current Overall Likelihood Per Frame = 5.26908373475453
Baum welch starting for 1 Gaussian(s), iteration: 2 (1 of 2)
bw Log File
I have looked through the log files but do not really understand what's happening there. When decoding I get the following error
MODULE: DECODE Decoding using models previously trained
Decoding 130 segments starting at 0 (part 1 of 1)
Could not find executable for %rootdir%\bin\pocketsphinx_batch at %rootdir%\scripts_pl\lib/SphinxTrain/Util.pm line 299.
Can't open %rootdir%/result/an4-1-1.match
SENTENCE ERROR: 100.000% (130/130)
The fact is that neither pocketsphinx.batch nor an4-1-1.match exist (however an4.match does exist).
Whatever help you could bring me will be welcome, thanks!!
Corisco
Now it's better. The main error string is:
Could not find executable for %rootdir%\bin\pocketsphinx_batch at %rootdir%\scripts_pl\lib/SphinxTrain/Util.pm line 299.
The decode process calls decode with test files and language model you are using and decoder produces match file. So you must:
1) compile pocketsphinx_batch
2) put it into %rootdir%/bin
>I have looked through the log files but do not really understand what's happening there.
Well, you can ask about details but not about everything at once. Start with something simple, for example with a single line in the log and ask questions about it.
It decodes!!! Thanks A LOT Nickolay!
However I still get some errors while training in modules
MODULE: 20 Training Context Independent models ........Phase 3: Forward-Backward
MODULE: 50 Training Context dependent ........Phase 3: Forward-Backward
All of them come from gauden.c
ERROR: "........\src\libs\libmodinv\gauden.c", line 1700: var (mgau= 0, feat= 2, density=34, component=0) < 0
ERROR: "........\src\libs\libmodinv\gauden.c", line 1700: var (mgau= 0, feat= 2, density=50, component=1) < 0
ERROR: "........\src\libs\libmodinv\gauden.c", line 1700: var (mgau= 0, feat= 2, density=73, component=2) < 0
I got several errors while compiling the pocketsphinx and sphinxbase nightly versions from svn (due to missed files: cache_lm.c lm.c and s2_mmap.c) so I am using releases 0.4.1 and 0.3 respectively. I am working with XP SP2 and VS 2005
Any clue?? Thanks
Corisco.
This error means you have not enough data to train senones. Just ignore it or use less senones or provide more data.
Ahmm! OK, I'll start testing some different models.
THANKS again for your help