Hi, Truecasing may lead to sparsity issues, but it alsp helps with name/noun distinctions in English (Henry Fisher, Henry the fisher) and noun/verb distinctions in German (verb wissen, noun Wissen).
-phi On Sat, May 5, 2012 at 1:57 PM, Panos Kanavos <[email protected]>wrote: > If I understand correctly, the source sentence should always be sent to > Moses > in its natural case, except for the first word which always has to be in > title > case? > > Also, does truecasing result in any data sparsity compared to the lowercase > approach? > > Thank you. > > Panos > > On Saturday 05 of May 2012 13:42:13 Philipp Koehn wrote: > > Hi, > > > > yes, you have to tokenize it and and possible change the case of > > a sentence starting word, if its natural casing is lowercased. > > > > -phi > > > > On Sat, May 5, 2012 at 4:27 AM, Panos Kanavos <[email protected]> > wrote: > > > Hi all, > > > > > > In a truecased model, do I have to do any special processing to the > input > > > sentence before sending it for translation to Moses? > > > > > > Thanks. > > > > > > Panos > > > _______________________________________________ > > > 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
