[Corpora-List] Evaluating Sentence Aligners

[Ruprecht von Waldenfels]

Dear Mike, 
from my experiments, I got very strange results if only one side was lemmatized/stemmed, i.e., in some cases, alignment quality improved, in others, it degraded.

A clear result, however, was that lemmatization helps alignment, at least with the aligners I tested, that is, bsa and hunalign; the latter proved more stable when presented with a text that was abridged. 

I think choosing aligners and whether or not to lemmatize depends on the task and the languages involved; some aligners need linguistic resources that might not be easily available. Also, if you can afford to throw away those parts that were aligned with low certainty, the picture changes again.


> > Are they results fairly insensitive to morphology?  I.e. does it matter
> > whether you stem one or both sides?
> >
> >    Mike Maxwell
> >    CASL/ U MD
> >
> >

-- 
Psssst! Schon vom neuen GMX MultiMessenger gehört?
Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger

_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora