Measuring Coverage of a Valency Lexicon using Full Syntactic Analysis

Recent development showed that valency information provides a great benefit in many areas of natural language processing. Building valency lexicons is however a complex and time-consuming task from both theoretical and practical points of view, since designing of the lexicon plays a crucial role in its future usability as well as its careful and considerated preparation. As for any manually created resource, it is complicated to evaluate its quality. In this paper we consider the usage of the syntactic parser synt for estimating the coverage of the Verbalex verb valency lexicon for Czech. For this task we extended the phrase extraction functionality of the parser, which we describe briefly. Finally we discuss our results and further development.