expanding an old language model

Speech Recognition Toolkit

Brought to you by: air, arthchan2003, awb, bhiksha, and 5 others

This project can now be found here.

expanding an old language model

Forum: Help

Creator: sachin singh

Created: 2012-08-23

Updated: 2012-09-22

sachin singh - 2012-08-23

hey friend,

i trying to use sphinx in my application. i want to use it for language
"Hindi", which is an Indian language. i found an opensource model from :- htt
p://sourceforge.net/projects/hindiasr/files/ .

firstly, i tried to use this model with sphinx 3 using "sphinx_decode". then i got stuck in how to add new words in prebuild model?. i tried to copy new words in ".dic" file with their phenom, but that doesn't work. i searched a lot on internet about how to expand default model, but nothing found.

then i switched to sphinx 4, which provide "java API". so i start building in Eclipse, i build "hello world" app successfully. then i tried to use "Hindi" model, but i stuck in configuration of "linguist model" in sphinx 4. because a grammar file is necessary in sphinx 4. but in "Hindi" model, no grammar file was present.

so after spending two weeks, i am nowhere. now i am thinking to use pocket
sphinx, with "Hindi" model, so is it right step?
most important question for me is -> how to add new word in an old model? if
you can download and see what resources are available in "hindi", that would
be great.
i had see an video on YouTube on adding words to dictionary, but they are
using ".jar file" and file they are using to add new words is not available in
my "hindi" model. i am confused and need a solution ASAP.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

bic-user - 2012-08-27

tried to copy new words in ".dic" file with their phenom, but that doesn't
work

You also need to add the words to language model. You need to create your own
language model http://cmusphinx.sourceforge.net/wiki/tutoriallm and merge it with available
(hindi.arpa). Seems it's possible with SRILM. google for that toolkit.

because a grammar file is necessary in sphinx 4. but in "Hindi" model, no
grammar file was present.

Read tutorial above carefully.

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

sachin singh - 2012-08-31

i switched to pocket sphinx. but now i am not able to test
pocketsphinx_continuous, it crashes after :- "Reading HMM transition
probability matrices hub4opensrc.cd_cu/transition_matrices" ....

also i was thinking to train an acoustic model but i didnt found
pocketsphinx_mdef_convert to convert binary to text .... but it gives me
linking error , compiler is not able to find definition of function's. what
should i do? ...

Error 5 error LNK1120: 4 unresolved externals
C:\sphinx\pocketsphinx\pocketsphinx-0.7\Debug\test_mdef.exe test_mdef
Error 2 error LNK2019: unresolved external symbol _bin_mdef_read referenced in
function _main
C:\sphinx\pocketsphinx\pocketsphinx-0.7\test_mdef\mdef_convert.obj test_mdef
Error 4 error LNK2019: unresolved external symbol _bin_mdef_read_text
referenced in function _main
C:\sphinx\pocketsphinx\pocketsphinx-0.7\test_mdef\mdef_convert.obj test_mdef
Error 3 error LNK2019: unresolved external symbol _bin_mdef_write referenced
in function _main
C:\sphinx\pocketsphinx\pocketsphinx-0.7\test_mdef\mdef_convert.obj test_mdef
Error 1 error LNK2019: unresolved external symbol _bin_mdef_write_text
referenced in function _main
C:\sphinx\pocketsphinx\pocketsphinx-0.7\test_mdef\mdef_convert.obj test_mdef

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Log in to post a comment.