Grounding the lexical semantics of verbs in visual perception using force dynamics and event logic

摘要: This paper presents an implemented system for recognizing the occurrence of events described by simple spatial-motion verbs in short image sequences. The semantics these is specified with event-logic expressions that describe changes state force-dynamic relations between participants event. An efficient finite representation introduced infinite sets intervals occur when describing liquid and semi-liquid events. Additionally, procedure using this presented inferring occurrences compound events, expressions, from primitive Using force dynamics event logic to specify lexical allows be more robust than prior systems based on motion profile.