Extracting an Ontology of Portrayable Objects from Wordnet