In this note, I illustrate the methodology we are currently using to acquire typed predicate-argument structures from corpora, with the aim of compiling a repository of corpus-based patterns for Italian verbs and obtaining an empirically sound inventory of argument type shiftings in context for linguistic research and NLP applications. The note is organized as follows. I first introduce the resource, then focus on the annotation of type mismatches between argument fillers and verb selectional requirements and their linguistic classification, based on Generative Lexicon Theory (Pustejovsky et al. 2008). Finally, I outline the ongoing attempt to combine typing annotation with standard coarse-grained (high-level) thematic role information (Bonial et al. 2011a, 2011b) carried out in collaboration with Senso Comune (Vetere et al. 2011). A discussion of ongoing evaluation and improvements follows.
Acquiring predicate-argument structures from corpora
JEZEK, ELISABETTA
2012-01-01
Abstract
In this note, I illustrate the methodology we are currently using to acquire typed predicate-argument structures from corpora, with the aim of compiling a repository of corpus-based patterns for Italian verbs and obtaining an empirically sound inventory of argument type shiftings in context for linguistic research and NLP applications. The note is organized as follows. I first introduce the resource, then focus on the annotation of type mismatches between argument fillers and verb selectional requirements and their linguistic classification, based on Generative Lexicon Theory (Pustejovsky et al. 2008). Finally, I outline the ongoing attempt to combine typing annotation with standard coarse-grained (high-level) thematic role information (Bonial et al. 2011a, 2011b) carried out in collaboration with Senso Comune (Vetere et al. 2011). A discussion of ongoing evaluation and improvements follows.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
