[Corpora-List] Evaluating Sentence Aligners

[Eric Garbin]

Dear Olivier,

Thanks for your thorough reply!  Have you compared the performance of
Alinea against a baseline in which sentences are aligned using just a
comparison of length (in words or in characters)? 

 It makes sense that sentence length would be useful when combined with
language specific features.  One can't depend on it alone, of course,
when the alphabets are too different.  Have you looked at how useful
sentence length is for evaluating Alinea on distant language pairs?  In
other words, if Alinea were to ignore sentence length completely, I
wonder how much of a hit you'd take on, say, French-Arabic precision and
recall on the same data.

--Eric Garbin


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