Hi,

Even I have approached the same way. I have repeated entries of the
dictionary while appending them in the training corpus. The AER reduced
roughly from 52. to 48. after repeating for 7 times and increased to 49
after 10 times.

Regards,
Murali Krishna Emani.


On Nov 28, 2007 5:19 PM, Ondrej Bojar <[EMAIL PROTECTED]> wrote:

> Hi,
>
> I don't use the dictionary file, but my guess is that all GIZA++ does is
> attach
> an extra portion to the input parallel corpus, pretending that each phrase
> in
> the dictionary is a sentence as such:
>
> cat    chat
> black  noir
> ...
>
> Giving it a weight is essentially just repeating that part several times.
>
> cat    chat
> cat    chat
> black  noir
> black  noir
> ...
>
> Ondrej.
>
> emani murali wrote:
> > Hello,
> >
> > I could hardly find any material as how to give weights to dictionary
> > file that is used in GIZA++. I would appreciate if anyone let me know
> > if there is any command line option that does so or any other method
> > for the same.
> >
> >
> > Regards,
> > Murali Krishna EMANI
> > IIIT Bangalore.
> > _______________________________________________
> > Moses-support mailing list
> > [email protected]
> > http://mailman.mit.edu/mailman/listinfo/moses-support
>
> --
> Ondrej Bojar (mailto:[EMAIL PROTECTED] / [EMAIL PROTECTED])
> http://www.cuni.cz/~obo <http://www.cuni.cz/%7Eobo>
>



-- 
Murali Krishna EMANI
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to