(sorry, apparently the email wasn't properly received by the list)
Hi,
1) assuming that you signed the French treebank license (google "anne
abeillé+French treebank"), getting the french treebank is easy
2) getting a version suitable for stat. dependency parsing is also easy:
http://www.linguist.univ-paris-diderot.fr/~mcandito/Rech/FTBDeps/index.html
3) you could also contact Marie Candito, Pascal Denis or Benoit Crabbé
([email protected],[email protected], [email protected]
) to get either a parser trained on a dependency version of the FTB or
the training data..
I assume that you don't read French but a dependency parser for
french was described in (candito et al, 2009)
http://www.linguist.univ-paris-diderot.fr/~mcandito/Publications/candito-crabbe-denis-guerin-taln2009.pdf
In short, the authors used a state-of-the-art constituency parser
(petrov et al, 06) + a maxent functionnal labeler to get dependencies
parses.
Now, there is an LREC upcoming paper by the same gang where they
compared this approach (although highly optimized using word
clustering for training) against the Ryan Mc Donald's dependency
parser trained on the French Dependency Bank.
Just contact them, they will be glad to help you.
All the best,
Djamé
_______________________________________________
Mt-list mailing list
