Hi, Not entirely sure what you are asking for here.
If you just want to have a word alignment between source and target, why not use GIZA++? If you do that, make sure to also throw in (parts of) the parallel corpus to get better alignment quality. If you want to have the implicit word alignment of Moses when translating the source to the reference, then you are faced with the problem that the models may not even allow the generation of the reference as a translation (think of an unknown word in the source). -phi On Tue, Jun 4, 2013 at 12:03 PM, f.fancellu <[email protected]> wrote: > Hi, > I have another question regarding word alignment information during > decoding. Is it possible for Moses to output word alignment information > between source and reference set during decoding? I can output a.wa files > containing word alignment information between a the test source side and the > output translation but I could find any files containing word alignment > information between test source side and test reference. > > Thanks! > F. > > _______________________________________________ > 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
