[Corpora-List] Processing Latin texts

[Joakim Nivre]

On Tue, 15 Apr 2008, maxwell at umiacs.umd.edu wrote:

> Joakim Nivre wrote:
> > We have trained and evaluated MaltParser (an open-source data-driven
> > dependency parser available from http://w3.msi.vxu.se/jha/maltparser/) on
> 
> (the above link seems to be broken)

Sorry, the link should be http://w3.msi.vxu.se/~jha/maltparser/
 
> > the Latin Dependency Treebank described by Bamman and Crane in their paper
> > at the 2006 Workshop on Treebanks and Linguistic Theories
> > (http://ufal.mff.cuni.cz/tlt2006/pdf/110.pdf). However, the results so far
> > are not very impressive because of the limited amount of training data and
> > the high amount of non-projective structures (especially in poetry).
> > Moreover, we did not use a part-of-speech tagger for preprocessing but
> > simply used the manual part-of-speech annotation in the treebank as input.
> 
> I'm unclear how you handled the POS tags--did both training and test data
> have POS tags?

Yes. 

Joakim
 
>    Mike Maxwell
>    CASL/ U MD
> 
> 
> _______________________________________________
> Corpora mailing list
> Corpora at uib.no
> http://mailman.uib.no/listinfo/corpora
> 

==================================================================
Joakim Nivre

Växjö University		Uppsala University
School of Mathematics		Department of Linguistics
and Systems Engineering		and Philology
SE-35195 Växjö			Box 635, SE-75126 Uppsala

Tel: +46 470 708992		Tel: +46 18 4717009
Fax: +46 470 84004		Fax: +46 18 4711094
E-mail: nivre at msi.vxu.se	E-mail: joakim.nivre at lingfil.uu.se

URL: http://www.msi.vxu.se/users/nivre
==================================================================
-------------- next part --------------
_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora