Hi, typically a tuning set should be a small part of the parallel corpus that is closest to the desired target domain, and it cannot be used in training the translation model or language model.
Tuning is not strictly necessary, sometimes the default weights give decent results, but more likely than not you will see significant improvements. -phi On Tue, Jul 16, 2013 at 6:55 PM, Heidi Heweidy <[email protected]> wrote: > I need to tune my arabic, anyone knows where I can get parallel data for > that? Also, is tuning necessary? > > _______________________________________________ > 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
