Contributo in volume, 2013, ENG

Domain Adaptation for Dependency Parsing at EVALITA 2011

F. Dell'Orletta and S. Marchi and S. Montemagni and G. Venturi and T. Agnoloni and E. Francesconi

ILC-CNR, ITTIG-CNR

The domain adaptation task was aimed at investigating techniques for adapting state-of-the-art dependency parsing systems to new domains. Both the language dealt with, i.e. Italian, and the target do- main, namely the legal domain, represent two main novelties of the task organised at Evalita 2011 with respect to previous domain adaptation ini- tiatives. In this paper, we define the task and describe how the datasets were created from different resources. In addition, we characterize the different approaches of the participating systems, report the test results, and provide a first analysis of these results.

Keywords

Dependency Parsing, Domain Adaptation, Self-training, Active Learning, Legal-NLP

CNR authors

Venturi Giulia, Montemagni Simonetta, Marchi Simone, Francesconi Enrico, Agnoloni Tommaso, Dell Orletta Felice

CNR institutes

ILC – Istituto di linguistica computazionale "Antonio Zampolli", ITTIG – Istituto di teoria e tecniche dell'informazione giuridica, IGSG – Istituto di Informatica Giuridica e Sistemi Giudiziari