> > You can use the --giza-option to tell GIZA to use a different model, > > but I don't believe the moses scripts can be told look for the > > nonstandard file that GIZA will generate in this case (they are named > > differently for the kind of model used). But, this wouldn't be hard > > to make this configurable or to hard code it. > > Doesn't choosing only one iteration for the subsequent models > essentially do this? We usually run with: > > -giza-option="m1=5,m2=0,mh=5,m3=1,m4=1" > > which for some reason I had internalized as simply "transferring" the > final HMM settings to models 3 and then 4, without any real re- > estimation. Am I hallucinating this? :)
I'm fairly sure it reruns one optimization iteration. And, in any case, it would use the HMM parameters to run model 4 alignment (with the heuristic initialization of the fertility and distortion parameters that GIZA uses), which is probably not particularly good. Chris _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
