README

Part-of-speech tagger for French.

Benoit Favre benoit.favre@lif.univ-mrs.fr 2012

Requires CRFSuite 0.12 (http://www.chokkan.org/software/crfsuite/) and python2 or pypy. crfutils.py and add-contextual-features.py (pos.py) are copied from the CRFSuite source code.

echo Ça le fait. | ./run-tagger.sh Ça PRO le CLO fait V . PONCT

Trained on the French Treebank 6 (http://www.llf.cnrs.fr/Gens/Abeille/French-Treebank-fr.php) in CoNLL09 format Word-tag association features are from the Lefff dictionnary (http://alpage.inria.fr/~sagot/lefff-en.html)

$ crfsuite tag -t -q -m ftb6.model ftb6_3.conll09.crfsuite Item accuracy: 35484 / 36340 (0.9764) Instance accuracy: 678 / 1235 (0.5490) Elapsed time: 2.230000 [sec] (553.8 [instance/sec])