Qualitative Semantic Representations

This paper reports on our work in the Transregional Collaborative Research Center Spatial Cognition (SFB/TR8). The aim of this research is to achieve a natural and effective dialogic interaction between a human and an intelligent robot on spatial navigation tasks. Therefore, we have introduced a conceptual level representation of human spatial knowledge on top of the robot’s quantitative representation. Specifically, we have developed a kind of graph, i.e., the Conceptual Route Graph, which combines the concepts of route graphs and qualitative spatial orientation calculi. After integrating the Conceptual Route Graph into a dialogue system, it shows several advantages. First, the Conceptual Route Graph can serve as the semantic framework for communication between the different components of a dialogue system. Second, spatial reasoning can be used to detect orientation mismatches in human route instructions, thus clarification subdialogues can be generated automatically. Moreover, the Conceptual Route Graph provides a formal interface to a robot for carrying out spatial navigation tasks.