Using Object Influence Areas to Quantitatively Deal with Neighborhood and Perception in Route Descriptions

Abstract. In the GRAAD project we are developing a knowledgebased system able of determine routes in a simulated urban environment and to generate natural language descriptions which are not distinguishable from those produced by human beings in similar conditions. In this paper, we present a new spatial model whose topology is based on the notion of object’s influence areas. An influence area is a portion of space that people mentally build around spatial objects to take into account neighborhood. We use this notion to formally define the properties of neighborhood, orientation and distance in a qualitative way. We also introduce the notion of an object's perception area, an area gathering all the locations from which an object can be perceived. Based on these notions, we describe two modules of the GRAAD System that are able to find routes in a simulated urban environment and to generate route descriptions in natural language which are analog to those created by people.

• Type: Conference
• Year: 2000
• Where: AI
• Authors: Bernard Moulin, Driss Kettani, Benjamin Gauthier, Walid Chaker
• Keywords: AI 2000, Artificial Intelligence, Influence Area, Natural Language, Simulated Urban Environment