Situating Spatial Templates for Human-Robot Interaction

Through empirical validation and computational application, template-based models of situated spatial term meaning have proven their usefulness to human-robot dialogue, but we argue in this paper that important contextual features are being ignored; resulting in over-generalization and failure to account for actual usage in situated context. Such a fact is significant to human-robot dialogue in that it constrains the manner in which we create interactive systems which can discuss their own physical actions and surroundings. To this end, in this paper we describe a study which we conducted to determine how acceptability ratings for spatial term meaning altered for oblique landmark orientations. Results demonstrated that spatial term meaning was indeed altered by interlocutor perspective in a way not predicted by current approaches to spatial term semantics.