Hi, if you want to ignore all other costs, then yes, you can set their weights to zero. I am not sure what exactly you are planning, but if you allow alternative phrase translations, the decoder is not able to limit them properly (by default only the 20 per input span are used) and you may end up with a lot of hypotheses with the same score, which may cause problems with pruning during search.
-phi On Wed, Sep 19, 2012 at 8:43 PM, Neda Noormohammadi <[email protected]> wrote: > Hi Moses, > > I am planning to generate the Moses output based on just reordering model. > In this task, the other models are not significant and the 1 best output is > a hypothesis with the highest reordering score. In fact, I want that the > outputs rank based on reordering model score and not the combination of all > models. Also, In this experiment the lexicalize reordering model is used. > > Is it sufficient that the weights of the other models are assigned to zero? > For this task, what do I do? > > Thanks > > _______________________________________________ > 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
