hi hala
On 22/06/2010 16:34, Hala Maghout wrote:
Hi,
I'm now working on building syntax based model using Moses. I'm
basically working on SAMT4 relaxation scheme. But I still have some
unclear points regarding rule extraction and scoring. I will formalize
those points into the following questions:
* Is SAMT4 relaxation scheme in Moses build a model equivalent to
the SAMT model specified in the paper
* :(/Andreas Zollmann and Ashish Venugopal. 2006.// //Syntax
Augmented Machine Translation via Chart Parsing//. In// //NAACL
2006 - Workshop on statistical machine translation//, New York.
June 4-9.). T/o be more specific does the rule scoring using
SAMT4 scheme take into account the probability of the rule given
its left hand side label ?or is the rule scoring method the
same as the method suggested by (Chiang, 2005) ?
it does take the LHS into account. eg. these 2 rules can have different
scores
NP -> book ||| livre
VP -> book ||| livre
* Does Moses do Rule generalization described in the same paper
above which adds for each extracted syntactic rule a Hiero style
rule by replacing all the nonterminals in the syntactic rule
with one non-terminal?
not sure i understand this question. If you mean are the syntax rules
automically converted to hiero rules, then no it doesn't
Many thanks,
Hala
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
