DepPattern: A Grammar Based Generator of Multilingual Parsers

DepPattern toolkit is a linguistic package providing a grammar compiler, PoS taggers, and dependency-based parsers for several languages.

The compiler, named compi-beta, was implemented in Ruby. It generates parsers in PERL from DepPattern grammars. To write formal grammars using the formalism "DepPattern", please, look up the tutorial.

In addition, DepPattern is provided with parsers for 5 languages: English, Spanish, Galician, French, and Portuguese. The parsers were implemented in PERL and are stored in the directory ``parsers''. Their ancestor is Multilingua, a more basic parser without morphological features. The parsers take as input PoS tagged text. To tag texts, DepPattern uses either Tree-Tagger or Freeling. Treetagger is provided by the distributed package. Freeling must be previously installed. In the directory ``parsers'', there is also a test parser, which was generated by compi-beta from a test grammar, stored in the ``grammars'' directory.



Subsections

Pablo Gamallo 2009-10-02