Hi Barry, Thank you for pointing out the location of the logic. I understand that we can safely discard the worse hypothesis if the 1-best case.
However, if we seek n-best tranlations, it would be problematic to recombine different translations and make the score of worse hypothesis same as the better one. For example, as in the following slide (p.24): http://www.inf.ed.ac.uk/teaching/courses/mt/lectures/phrase-decoding.pdf "did not give" from "Joe" is the worse hypothesis than "did not give" from "Mary". After hypothesis recombination, the path "Joe did not give" can survive even though we have a better hypothesis that covers 3 foreign words. What if it is worse than "Mary did not give" and pruned out for this reason? I'm not sure it would be problematic to recombine identical translations if we seek n-best translatins. Does it lead too many identical translations in the n-best list? Best regards, -- Hwidong Na <[email protected]> KLE lab, POSTECH, KOREA 2011-07-05 (화), 08:51 +0100, Barry Haddow: > Hi Hwidong > > Yes, if you're extracting nbest lists, then you have to maintain backpointers > to the recombined hypothesis. You can see the logic for this in the > AddPrune() method in HypothesisStackNormal.cpp. > > If you're only seeking the 1-best, then it's possible to discard the lower > scoring of two recombined hypothesis, because you know for sure it can never > form part of the 1-best, > > best regards - Barry > > On Tuesday 05 July 2011 07:51, Hwidong Na wrote: > > Hi, > > > > The statistical machine translation book explains hypothesis > > recombination as in the following slide. > > > > http://www.statmt.org/book/slides/06-decoding.pdf > > > > I wonder how can we backtrace recombined hypothesis after hypothesis > > expansion is terminated. It seems fine when we recombine identical > > English output because all backtraces give identical translations. > > However, if we recombine different translations, which backpointer do we > > follow? If we read off n-best translations, do we follow all of > > different backpointers? > _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
