Lexical acquisition for computational grammars

In this dissertation, we focused on the problem of insufficient lexical coverage of computational grammars that rely on large, hand-crafted lexicons. As a possible solution to this problem, we proposed a novel lexical acquisition (LA) technique. This technique enables the automated acquisition of lexical entries for words which are not listed in the lexicon of a given grammar. We started with the description of popular, state-of-the-art implementation paradigms for computational grammars of natural language. Chapter 2 described context-free grammars and two extensions thereof, namely probabilistic context-free grammars and attribute-value grammars. The focus of the discussion was on the latter type since the computational grammars we employed for the experiments presented in our work are attribute-value grammars. We then presented the two grammars which were used throughout this thesis, namely the Alpino grammar of Dutch and the GG grammar of German. We also underscored the importance of the lexicon in those grammars. It is the place where the words covered by the grammar are specified together with their morphological properties and combinatorial potential. Next, Chapter 3 outlined the problem investigated in this dissertation, namely the insufficient lexical coverage of grammars which rely on handcrafted lexicons. We discussed previous studies which showed that unknown words pose a serious problem for the parsing performance of such grammars.