Communicating with Navigation Systems about Places

The aim of this research is to investigate the initial phase of route communication with a navigation system: the specification of a route by an origin and a destination. Comparing inputting of a place description to the ways that people communicate about places with each other, we observe gaps impacting the usability and efficiency of navigation systems. From these gaps we identify future research questions. This research only looks at the communication about places, and leaves other phases of the route communication such as the provision of route directions phase and confirmation and closing phase for future work. 1 Problem and motivation In daily life, people frequently communicate to each other about place. Although they have varying and incomplete descriptive knowledge about their spatial environment, their communication regularly succeeds. Similarly, navigation systems provide interfaces for inputting place descriptions specifying an origin and a destination. A place description in the context of this research is an answer to a where question [1]. Current navigation systems reveal mismatches between their approaches and their users’ cognitive reasoning and verbalization of place descriptions. Firstly, systems use gazetteers for toponym resolution [2]. To ease the burden of natural language processing, system interfaces come with structured input formats. Users have to be adaptive enough to describe their intended places in systems’ format, e.g., postal addresses. In this given example, users cannot use the system if they do not know the addresses or their desired places have no formatted postal address. For example, a hospital occupying a block would not have an address with one street number and one street name. Additionally, gazetteers consist of a set of geographic placenames, each georeferenced with a pair of coordinates. This way of georeferencing is not a realistic representation of our world: geographic features have a spatial extent and should be represented by regions rather than a point at the centre position. Also, gazetteers do not store relations between places. Secondly, systems have no awareness that their gazetteers are incomplete and outdated, and the structure and extent of gazetteers are implicit to users. When users look for places names that do not exist in the gazetteers, the systems might offer a close match but would never locate the desired places precisely no matter how users amend input. Contrary to formatted input in systems, humans describe places in a flexible manner. Winter and Wu [3] studied the route communication with a state-of-theart navigation system, Metlink’s Journey Planner, and observed several gaps in communicating with the system. This research looks at state of the art navigation systems, and identifies fundamental questions relating to the cognitive base of communicating with systems about place in general for future research. The communication between users and route planning systems consists of three phases: the initial phase where users ask information for directions, the second phase where the systems offer information, and the last phase of confirmation and closing [4]. This research focuses on the initial phase, and proposes a system that allows users’ input in a flexible manner as they use in daily life. The hypothesis of this research is that current systems have fundamental gaps in their understanding people’s descriptions of place.