In this paper we investigate if conceptual information concerning objects' affordances as possibilities for actions anchored to an object can be at least partially acquired through language. Considering verb-noun pairs as the linguistic realizations of relations between actions performed by an agent and objects we collect this information from the ImagAct dataset, a linguistic resource obtained from manual annotation of basic action verbs, and from a web corpus(itTenTen). The notion of affordance verb as the most distinctive verb in ImagAct enables a comparison with distributional data that reveal how lemmas ranking based on a semantic association measure that mirror that of affordances as the most distinctive actions an object can be involved in.
(Fore)seeing actions in objects. Acquiring distinctive affordances from language
Irene Russo;Irene De Felice;Francesca Frontini;Monica Monachini
2013
Abstract
In this paper we investigate if conceptual information concerning objects' affordances as possibilities for actions anchored to an object can be at least partially acquired through language. Considering verb-noun pairs as the linguistic realizations of relations between actions performed by an agent and objects we collect this information from the ImagAct dataset, a linguistic resource obtained from manual annotation of basic action verbs, and from a web corpus(itTenTen). The notion of affordance verb as the most distinctive verb in ImagAct enables a comparison with distributional data that reveal how lemmas ranking based on a semantic association measure that mirror that of affordances as the most distinctive actions an object can be involved in.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
