I think word alignment output in the decoder has been disabled because of some memory issues I don't understand in detail. What I usually do nowadays is putting the phrase-internal word alignments in an additional factor on the TL side of the translation table. In conjunction with the phrase alignments that you get from -report-segmentation, this allows you to reconstruct the word alignments relatively easily.
/Christian On Fri, 02 Jul 2010, [email protected] wrote: > Hi all, > I'd really need to know exact word-to-word alignment, but I have not > managed yet moses to print out his information: > My binary phrase table files are compiled with the > -alignment-info. And, then, I call moses with the -use-alignment-info > and -print-alignment-info. As well as verbose (-v *) , and trace (-T > ...) options. > Nevertheless, I'm still not able to get word-to-word alignments > information. > Am I doing anything wrong? > Or, in case that moses does not support decoding with w2w info > anymore... Is there any other way to get this information, even no > matter whether decoding would get slower? > > Thanks, > Isaias > > > > ---------------------------------------------------------------- > This message was sent using IMP, the Internet Messaging Program. > > > _______________________________________________ > 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
