Hi Sanne, The files lex.f2n and lex.n2f follow the format A B Pr(A | B) where Pr(A|B) is the probability of translating from B to A, computed from GIZA alignments as Pr(A | B) = count(A aligned B) / total count B Note that: A, B could take the value NULL, and a word is considered to be aligned to NULL, if it has no GIZA alignment w.r.t a sentence pair.
For lex.f2n, B is in the source (foreign) language, while A is in the target (english) language. Vice versa for lex.n2f. Hope that helps, Cheers, Thang On Wed, May 6, 2009 at 9:27 PM, Sanne Korzec <[email protected]> wrote: > Hi, > > > > Could someone explain to me what is stored in the lex.f2n and lex.n2f > files? And how this is computed? > > > > Thanks > > Sanne > > _______________________________________________ > Moses-support mailing list > [email protected] > http://mailman.mit.edu/mailman/listinfo/moses-support > > -- Luong Minh Thang WING group, School of Computing, National University of Singapore http://wing.comp.nus.edu.sg/~lmthang
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
