A Bootstrapping Approach to Parser Development
Izaskun Aldezabal, Koldo Gojenola, Kepa Sarasola · 2000
This paper presents a robust parsing system for unrestricted Basque texts. It analyzes a sentence in two stages: a unification-based parser builds basic syntactic units such as NPs, PPs, and sentential complements, while a finite-state parser performs syntactic disambiguation and filtering of the results. The system has been applied to the acquisition of verbal subcategorization information, obtaining 66% recall and 87% precision in the determination of verb subcategorization instances. This information will be later incorporated to the parser, in order to improve its performance. 1 Introduction As NLP-based applications are growing, there is a stronger need for wide-coverage parsing systems. At the moment, comprehensive grammars are available for some languages, like English [Briscoe and Carroll 1993], or parallel LFG German, French and English grammars [Butt et al., 1999], developed after a considerable effort. Moreover, the huge size of the now available corpora demands succe...