Hi, sure - you could just remove low scoring phrases according to the direct phrase translation probability, keep only the n-best, and so on.
Check for instance: Howard Johnson, Joel Martin, George Foster, and Roland Kuhn (2007). Improving Translation Quality by Discarding most of the Phrasetable. EMNLP. -phi On Tue, Mar 22, 2011 at 12:29 PM, Mehmet Tatlıcıoğlu <[email protected]> wrote: > Hi, > I am working on pruning the translation table, and I was wondering if > pruning is possible without having bitext data. Because, I have only > translation table file, namely phrase-table.gz, and I'd like to prune it > with a threshold value. I took a look at the manual, and what it says that I > need the parallel text for the pruning. Is that possible to prune it without > source and target texts? > Greets. > > -- > Mehmet Tatlıcıoğlu > Cell Phone: +90 532 201 85 64 > > _______________________________________________ > 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
