Hi All,
I am Ibrahim from Sakhr company, I am working on the Arabic à English statistical machine translation of the company. I am using the Giza++ under linux for training , SRILM toolkit under linux for building language models , and moses decoder under windows with MSVS 2005 for decoding . I found that moses gives bad translation when I use binary language model instead of the ASCII one for the same English Corpus. By the way , I don’t modify the configuration file of moses it is ( 0 0 3 Language Model Path) for both language models , I tried ( 1 0 3 binary Language Model Path ) but moses said that this type of language model is not supported. Is this problem because of running moses under windows or what ? Thanks,
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
