CROATPAS: A Resource of Corpus-derived Typed Predicate Argument Structures for Croatian

CROATPAS is a digital language resource for Croatian focusing on verbal semantics and able to frame linguistic phenomena – such as verbal polysemy and metonymic shifts – following the compositional approach of James Pustejovsky’s Generative Lexicon Theory and subsequent work. Just like its Italian sister resource TPAS - managed by Professor Elisabetta Jezek at the University of Pavia - the acronym CROATPAS stands for Croatian Typed Predicate Argument Structure resource: both resources are corpus-derived collections of verb argument structures whose argument slots have been tagged using a specific set of semantic labels called Semantic Types. CROATPAS is made up of four key-components, namely: 1) a representative corpus of contemporary Croatian; 2) a shallow ontology of Semantic Types; 3) a lexicographic methodology developed by Professor Patrick Hanks called Corpus Pattern Analysis (CPA); and 4) the adequate corpus tools (Sketch Engine and a tailor-made CPA editor). The potential uses and purposes of the resource range from multilingual pattern linking between compatible resources to computer-assisted language learning (CALL).