Convert Gigaword language Model to DMP

[Jagadeesh Balam]

Hi,

Has anybody successfully used one of the language models on Keith Vertanen's site with pocketsphinx? He metions that we need to resort the n-grams to do this successfully but I am a beginner and I have no idea what that exactly means. If you have successfully converted the .arpa to .DMP and got it working, please share the steps that I need to follow to do this.

Thanks,
Jagadeesh

[David Huggins-Daines]

Hi! Yes, there is a script which comes with SphinxBase called sphinx_lm_sort, which will put them in the right order:

dhuggins@slim:~$ sphinx_lm_sort -?
Usage: /usr/local/bin/sphinx_lm_sort < INPUT_LM > OUTPUT_LM

This program reads an ARPA-format language model and
sorts the N-Grams into the order required by Sphinx.
It also replaces 'missing' (actually nonexistent)
backoff weights with dummy values in order to satisfy
all versions of Sphinx.

[Jagadeesh Balam]

Hi David,

Thanks for the reply. I didn't know about this but I had a problem running it though. I tried it on both the 5k and 64k models, it just quits before completely sorting the file in the 1-grams. Has this happened before?

Thank you,
Jagadeesh

[Jagadeesh Balam]

It was some problem with my cygwin perl. I tried it on windows and it worked. Thanks again for your quick reply David.

J

[munvar ali]

Hi jagadeesh,

I appriciate ur work.

Can you pls tell me the procedure to run pocketsphinx in wince or any other mobile ??

i already worked in SPHINX-4 for indian accent english and achieved that......

please tell me the step by step procedure so that i can run pocketsphinx in mobile.

as we are doing research on pocketsphinx,by sharing our knowledge together we will achieve our goals

thank you inadvance.....