MaltParser at the EVALITA 2009 Dependency Parsing Task
Alberto Lavelli, Johan Hall, Jens Nilsson, Joakim Nivre · 2009
This paper describes our participation in the EVALITA 2009 Depen- dency Parsing Task with a version of MaltParser. Reusing feature models devel- oped in the CoNLL shared task 2007, we evaluated four different parsing algo- rithms implemented in MaltParser and found that the best results were achieved with Covington's non-projective parsing algorithm. In the final evaluation, our system finished third in the main task and second in the pilot task.