VoxForge
Hi
Julius man page says -
2-gram and reverse 3-gram language models are used. The Standard ARPA format is supported. In addition, a binary format N-gram is also supported for efficiency. The tool mkbingram can convert binary N-gram from the ARPA language models.
So I dont understand now if any n-gram can be used or only specifically 2 gram and reverse 3-gram.Since there are 3-gram language models available on CMU Sphinx page which can now be used for julius.
Also please give any pointers on where a language model for julius can be found. Thanks a ton
--- (Edited on 8/27/2010 11:26 am [GMT-0500] by Ravishanker) ---
Also, any pointers on text corpus for building language models is highly appreciated. Thanks :)
--- (Edited on 8/27/2010 12:19 pm [GMT-0500] by Ravishanker) ---
>Also please give any pointers on where a language model
>for julius can be found
See Keith vertanen's: English Gigaword language model training recipe
I don't think you need reverse 3-grams with the newest version on Julius.
Ken
--- (Edited on 8/29/2010 1:11 pm [GMT-0500] by Visitor) ---
Thanks ken :) Is there a corpus available for developing language models. For Keith's model is there a way to get the source text corpus. Any open source text corpus available for building language model? Thanks :)
--- (Edited on 8/31/2010 1:59 pm [GMT-0500] by Ravishanker) ---
>For Keith's model is there a way to get the source text corpus.
>Any open source text corpus available for building language model?
As listed on the VoxForge dev page:
DIY corpus using Search Engines (like Google)
--- (Edited on 9/11/2010 11:39 pm [GMT-0400] by kmaclean) ---