Reasoning About Actions in Narrative Understanding

Abstract

Reasoning about actions has been a focus of interest in AI from the beginning and continues to receive attention. But the range of situations considered has been rath er narrow and falls well short of what is n ee ded for understanding natural languag e. Language und e rstanding requires sophisticated reasoning about actions and events and the world 's languag es emp loy a variety of grammatical and lexical devices to construe, direct attention and focus on, and control inferences about actions and events. We impleme nt ed a neurally inspired computational mod el that is able to reason about linguistic action and eve nt descriptions , such as those found in ne ws stories. The system uses an active event represe ntation that also seems to provide natural and cognitively motivated solutions to classical problems in logical theories of reasoning about actions. For logical approaches to reasoning about actions, we suggest that looking at story understanding sets up fairly strong desiderata both in terms of th e fine-grained event and action distinctions and the kinds of real-time inferences required.