thanks

so the distance penalty + 6 * msd-bidirectional-fe features account for the
7 * weight-ds

but i can't see on
http://www.statmt.org/moses/?n=Moses.FAQ#ntoc7
nor
http://www.statmt.org/moses/?n=Moses.AdvancedFeatures
the correspondence between each weight-t and each feature phi(f phrase|e
phrase), lex(f phrase|e phrase), etc.

-dave

On 1 November 2010 15:14, Barry Haddow <[email protected]> wrote:

> Hi Dave
>
> For weight-t, see the moses faq
> http://www.statmt.org/moses/?n=Moses.FAQ#ntoc7
>
> For weight-d, see
> http://www.statmt.org/moses/?n=FactoredTraining.BuildReorderingModel
>
> The first weight-d is the distance penalty.
>
> cheers
> Barry
>
> On Monday 01 November 2010 15:03, David Matthews wrote:
> > I've run experiment.perl and got out what seem to be moses 14 default
> > features (and weights)
> >
> > Can someone please tell what translation and distortion weights
> corresponds
> > to what features? i.e. which feature is weighed by the 1st, 2nd, ... etc
> > weight-t and weight-d?
> >
> > I had a look in the docs and a quick search in the archive but couldn't
> > turn up anything
> >
> > cheers,
> >
> > dave
> >
> > # distortion (reordering) weight
> > [weight-d]
> > 0.0706199
> > 0.0798381
> > 0.0395168
> > 0.0641386
> > 0.113956
> > 0.00364072
> > 0.0736139
> >
> > # language model weights
> > [weight-l]
> > 0.0738448
> >
> > # translation model weights
> > [weight-t]
> > 0.0716699
> > 0.0195452
> > 0.0625724
> > 0.0532344
> > 0.171888
> >
> > # no generation models, no weight-generation section
> >
> > # word penalty
> > [weight-w]
> > -0.101921
> >
> > [distortion-limit]
> > 6
>
> --
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
>


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

Reply via email to