Hi, the binarization that Huang et al are talking about takes place after grammar extraction, and could be done on-demand just before decoding.
Moses is not doing this, it is doing cube pruning over rules of arbitrary arity. It is worth investigating, if the Huang et al binarization is more search-efficient. Let me know, if you are interested in implementing this, I could give you some pointers. -phi On Fri, Apr 8, 2011 at 6:06 PM, Thu Vuong Hoai <[email protected]> wrote: > Hi. > I've found this paper <link>, so that, I want to know that rule table which > we have after training process is binary synchronous SCFG or not? > > -- > Thu. > > _______________________________________________ > 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
