Hi Fabienne!

Yes, you can find an implementation of GHKM minimal rule extraction (Galley
et al., 2004) by Philip Williams in this dir:

moses/scripts/training/phrase-extract/extract-ghkm

Philip does not seem to have provided a documentation in this dir. Yet you
can just run make to generate the executable and look into extract-ghkm.cpp
for the file requirements. If you try it on Moses, since it currently only
extracts minimal rules and not composed rules, you may not get fancy BLEU
scores.

Alternatively, Liang Huang and Haitao Mi have a python implementation in
assembla, which extracts both minimal and composed rules. And you can google
it out. Pre- and post-processing scripts are needed since it accepts and
outputs files differently from Moses does. If you are interested, I can send
you my scripts so that you don't need to reinvent the wheel.

We are now experimenting with different rule extraction and scoring
strategies on French-English translation and Chinese-English translation
(Robert and me). If you are interested, let us discuss in more detail.


Cheers,

Yang Gao






On Tue, Jul 5, 2011 at 3:30 PM, Fabienne Braune
<[email protected]>wrote:

> Hello,
>
> ghkm seems to be implemented for moses. Is there any documentation or
> tutorial about this implementation?
>
> Bests,
> Fabienne
>
> _______________________________________________
> Moses-support mailing list
> [email protected]
> http://mailman.mit.edu/mailman/listinfo/moses-support
>
>


-- 
Yang Gao (Vivian)
----------------------------------------
Webpage:  http://homepages.inf.ed.ac.uk/ygao2
Research Associate, School of Informatics, University of Edinburgh
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to