PISA: A measure of Preference In Selection of Arguments to model verb argument recoverability
Giulia Cappelli, Alessandro Lenci · CINECA IRIS Institutial research information system (University of Pisa) · 2020
Our paper offers a computational model ofthe semantic recoverability of verb arguments,tested in particular on direct objects and In-struments. Our fully distributional model isintended to improve on older taxonomy-basedmodels, which require a lexicon in addition tothe training corpus. We computed the selec-tional preferences of 99 transitive verbs and173 Instrument verbs as the mean value of thepairwise cosine similarity between their argu-ments (a weighted mean between all the argu-ments, or an unweighted mean with the top-mostkarguments).Results show that ourmodel can predict the recoverability of objectsand Instruments, providing a similar result tothat of taxonomy-based models but at a muchcheaper computational cost.