Action Identification and Local Equivalence of Action Verbs: the Annotation Framework of the IMAGACT Ontology

Actions are productive concepts, but they are both linguistically and cognitively underdetermined: what defines an action in the event continuum is still an open question. The linguistic encoding of actions offers both problems and solutions to the issue of identifying these concepts. First of all, many action verbs do not identify one single action, but can refer to different action concepts. Secondly, each language categorizes actions in its own way. The IMAGACT Ontology of Action adopts a flexible approach to categorization that allows us to make a semantically coherent discrimination of action concepts across different languages. To this end IMAGACT employed the systematic annotation of Local Equivalence, i.e. the property that different verbs (with different meanings) can refer to the same action concept. However, Local Equivalences alone do not solve the problem of action identification: a further distinction of Local Equivalence relations is required in order to separate productive from non-productive equivalences. In fact, when such productivity is missing, Local Equivalences are not essential for action concept identification.