Hi All,

I am training a standard translation model using moses (using the
train-factored-phrase-model.perl
script). However, I want to use HMM model for alignments while running
GIZA++. I have the
following questions in this context:

1. What is the default alignment model used by GIZA++ when producing
alignments?
2. How to change the configuration to only use HMM based word alignments?

I found the following relevant parameters in the giza config file. However,
how should I set them to
make sure only HMM alignments are produced? Does a non-zero value of
"HMMIterations" mean
that it only produces HMM alignments? If so, do I need to make the other
model iterations equal to
zero?

Thanks in advance,

Thanks,
Manoj.

HMMITERATIONS: 5

MODEL1ITERATIONS: 5

MODEL2ITERATIONS: 0

MODEL3ITERATIONS: 3

MODEL4ITERATIONS:3

MODEL5ITERATIONS: 0
MODEL6ITERATIONS: 3
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to