Exploring Multilingual Semantic Role Labeling

This paper describes the multilingual semantic role labeling system of Computational Linguistics Group, Trinity College Dublin, for the CoNLL-2009 SRLonly closed shared task. The system consists of two cascaded components: one for disambiguating predicate word sense, and the other for identifying and classifying arguments. Supervised learning techniques are utilized in these two components. As each language has its unique characteristics, different parameters and strategies have to be taken for different languages, either for providing functions required by a language or for meeting the tight deadline. The system obtained labeled F1 69.26 averaging over seven languages (Catalan, Chinese, Czech, English, German, Japanese, and Spanish), which ranks the system fourth among the seven systems participating the SRLonly closed track.