Identification of the Initial Entity in Granular Route Directions

Current navigation services assume wayfinders new to an environment. In contrast, our focus is on route directions for wayfinders who are familiar with the environment, such as taxi drivers and couriers. We observe that people communicate route directions to these wayfinders in a hierarchical and granular manner, assuming some shared knowledge of the environment. These route directions do not focus on the route but rather on the destination. In this paper we solve the first problem of automatically generating route directions in a hierarchical and granular manner: finding the initial entity of such a communication. We propose a formal model to determine the initial entity, based on Grice’s conversational maxims, and applied to a topological hierarchy of elements of the city. An implementation of the model is tested for districts in a political subdivision hierarchy. The tests show a reasonable behavior for a local expert, and demonstrate the efficiency of granular route directions.