Zakieh, 

Look for a mis-match between the target half of the
parallel corpus and the language model data. Did you also apply POS
tagging to the language model? 

Tom 

On Tue, 27 Dec 2011 11:44:04
+0330, Zakieh Shakeri  wrote:  
Hi all,

I'm working on factored model
for English to Persian translation.I have POS tags for both source and
target languages. when I used only source language tags, the BLEU became
a little better(+0.5) but with applying tags for both sides, BLEU drops
drastically(-15).
I'll appreciate it if anyone can help me.

regards,

Zakieh 
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to