Hello everybody,

I wanted to understand how the word alignment is used while scoring phrases
and during translation.
I went through the link
http://www.statmt.org/moses/?n=FactoredTraining.ScorePhrases

It mentions that : " *An enhanced version of the scoring script outputs the
word-to-word alignments between f and e as they are in the files (extractand
extract.inv) generated in the previous training step **"Extract
Phrases*"<http://www.statmt.org/moses/?n=FactoredTraining.ExtractPhrases>
."

The translation tables generated by default look like this :

> grep '| in europe |' model/phrase-table | sort -nrk 7 -t\| | head
 in europa ||| in europe ||| 0.829007 0.207955 0.801493 0.492402 2.718

The web page seem to suggest that using this "enhanced script" we can get
the aligment points in the translation tables.

> grep '| in europe |' model/phrase-table | sort -nrk 7 -t\| | head

 in europa ||| in europe ||| (0) (1) ||| (0) (1) ||| 0.829007 0.207955

Which script is this link referring to?
Does this mean if my translation table does not have the alignment points
marked, no weightage is given to word alignments during translation?

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

Reply via email to