Elaborate Descriptive Information in Indoor Route Instructions

Elaborate Descriptive Information in Indoor Route Instructions Vivien Mast (viv@tzi.de) Cui Jian (ken@informatik.uni-bremen.de) Desislava Zhekova (zhekova@uni-bremen.de) I5-[DiaSpace], SFB/TR8 Spatial Cognition, University of Bremen Cartesium, Enrique-Schmidt-Strase 5, 28359 Bremen, Germany Abstract The following paper presents the enhancement of indoor route instructions with descriptive generation strategies. We con- sider the latter to be highly important for the quality and help- fulness of automatically generated indoor route instructions. We conducted an experiment showing that participants receiv- ing route instructions enriched with elaborate descriptive in- formation instead of step-by-step procedural information for crucial route segments performed better in objective and sub- jective measures than those receiving only basic prescriptive route instructions. Based on the gained knowledge, we con- clude that descriptive strategies are an important part of indoor route instructions and should be actively considered in system development. Keywords: indoor route instructions; descriptive information; wayfinding; spatial cognition; navigation Introduction Both navigation in indoor and outdoor environments profit from the use of landmarks since they are distinctive, eas- ily recognizable and highly memorable (Sorrows & Hirtle, 1999). Humans select landmarks for their distinguishing characteristics (Presson & Montello, 1988). Although the im- portance of landmarks in route instructions is well established (Allen, 1997; Denis, 1997; Richter, 2007; Raubal & Winter, 2002), most research in automatic generation of route instruc- tions focuses on one aspect of landmarks, namely to indicate the location at which a reorientation should take place in a network of paths. The main assumption of this approach is that good route instructions contain tightly coupled descrip- tive and prescriptive information. Therefore, current systems rely almost entirely on what Denis (1997) classified as Type 2 utterances – utterances coupling an action with a landmark. This leads to highly concise route instructions, but also limits the amount of descriptive information for each reorientation point to the mentioning of one landmark, possibly locating it with respect to the user. While this approach is particularly useful for car navigation (Brenner & Elias, 2003) which occurs in network space, i.e. along a street network where clearly identifiable nodes (inter- sections) are connected by edges (streets), in pedestrian nav- igation the case is different. Pedestrian navigation includes many areas that belong to scene space: open areas which are characterized by the absence of clearly identifiable nodes and edges (R¨uetschi, 2007; Schuldes et al., 2011). In net- work space, wayfinding consists mainly of selecting a path at each decision point, whereas in scene space, wayfinding is characterized by activities such as searching, exploring, and matching. There are no clear paths to choose from, but large spaces, where piloting between landmarks is necessary. Ori- ented search might be used if the expected landmark cannot be seen (Allen, 1999). In such areas, route graph represen- tations, and the resulting procedural information do not cor- respond very well to the needs of the wayfinder, as the func- tion of landmarks changes from identifying a turning point to more vague orientational aid. Indoor navigation has elements of both network and scene spaces. In addition, indoor spaces are characterized by a very limited amount of different land- mark types and a lack of highly salient landmarks. Usually landmarks consist mainly of doors, corridors and staircases, only very few of which are highly distinctive in comparison to outdoor landmarks which can be very diverse (a church, a petrol station, multiple intersections of different types, etc.). For this reason, the central roles of landmarks, i.e. signal- ing where actions should take place, as well as confirmation, are difficult to obtain in indoor scenarios. Additionally, this increases the difficulty of memorization, as it leads to instruc- tions which contain a series of highly similar utterances. A possible solution for these problems is the integration of more elaborate descriptive information into indoor route in- structions. This can be realised by basing instructions on a scene space representation of space, and using a descriptive strategy for generating route instructions for those areas that can be characterized as scene space: Instead of superimpos- ing abstract network representations onto open space areas, thereby producing a number of turning points and paths for an area which is viewed by a wayfinder as a coherent whole, this scene is described as one entity, and the location of the scene exit is described with respect to the scene. We assume that by introducing more elaborate descriptive information into in- door route instructions we can gain configurations of land- marks that can serve as highly salient landmarks, where sim- ple landmarks will yield no sufficient differentiation. More- over, we expect that the scene descriptions will enable more efficient localization of scene exits in the descriptions, mini- mizing the number of prescriptive statements. In contrast, the imposition of abstract networks onto open spaces will yield extra turns. We expect route instructions which integrate the descriptive approach to make it easier for participants to build up a mental image of the route in advance, leading to better memorization and increased confidence. In addition, mixing scene descriptions with prescriptive statements should yield more diverse route instructions, thereby additionally support-