Hi, I would say that a minimum should be a million words, but if the domain is limited, you may do well with less. I'd say, go ahead and try and check if you are happy with the quality. One issue is word alignment, so if you have a dictionary, it may help a lot.
-phi On Dec 10, 2007 5:17 PM, leonardo <[EMAIL PROTECTED]> wrote: > I am working in a project for translating some minority Mexican languages into > Spanish and vice versa. As you may have encountered when translating not > mainstream languages, it is the scarce resources for these languages. Is it > wise > to uses a statistical translator (like moses) when the corpus is small (= a > few > thousand lines, for example)? > Approximately, how many lines of training text can give us a decent > translation? > > > Thanks a lot for your answer! > > > _______________________________________________ > 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
