GIZA++ is a statistical machine translation toolkit that is used to train IBM Models 1-5 and an HMM word alignment model. This package also contains the source for the mkcls tool which generates the word classes necessary for training some of the alignment models.
hi, i want to know how to set the null-word probability in IBM Model 1, the probability of inserting a null after a source word. I have just seen the setting of IBM Model 3,
Hoping your response , thank you
hi, i want to know how to set the null-word probability in IBM Model 1, the probability of inserting a null after a source word. I have just seen the setting of IBM Model 3, Hoping your response , thank you