Conceptual Aspects: How Landmarks Can Be Described in Data Models - Landmarks: GIScience for Intelligent Services - page 111

location-based services literature personalization has been discussed. Services
learn from tracking and observing the behavior of a particular user with the goal
to providing tailored individual information rather than staying a neutral service
that provides the same information for everyone. Personalization is not limited
to mobile location-based services, of course. Search engines access a user's
search history in tailoring the response to their search request, which means that
different people do already get different responses on the same requests [ 67 ] .
Location-based services take a step further. Positioning technology on board of
smartphones allows tracking of visited places, and methods of data mining and
knowledge extraction from trajectories are now well known [ 70 ] .

￿ There will be objects in the database that stand out in some contexts, but not in

others. Thus, whatever landmarkness means it has to be considered together with
context parameters. A visually outstanding object in an environment may not be
known to a visually impaired person. The name or location of a bus stop may
not be known to a car driver. A café in a mall may be a good anchor for indoor
descriptions, but may be unsuited for car driving instructions, despite its address.
A phone booth may be a suited reference to mark a turn location for a pedestrian,
but may be too small in spatial granularity to safely guide a car driver. Context
matters, and must be part of a landmark model from the start.

￿ For different human communication partners different symbols or names when

referencing to landmarks may be appropriate. The first-time tourist may find a
reference to a type (“at the church”) more appropriate than the local, who might
prefer the name (“at St Francis”). Thus, context determines not only the choice
of the landmark, but also the way how to refer to it.

￿ As a consequence, a spatially intelligent system must be a context-aware system.

It does matter whether the conversation partner is driving a car, visually impaired,
sitting in a wheelchair, riding a bike, a public transport user, familiar with
the environment or not, and so on. In each context some references to land-
marks are more appropriate than other references. Going back to Janelle's [ 28 ]
categorization of communication constraints (Table 3.1 ) , a system architecture
typically assumes a communication situation as synchronous and non-co-located.
It may be aware of the human communication partner's location (as location-
based services are) but it has no further sensors to explore and adapt to the
communication situation. Since making a system context-aware is such a hard
task (compared to ease with which people attend to capturing and considering
context) the usual loophole of system designers is to devise a specialized system
for each context. We call this solution the antonym of a spatially intelligent
service.

With landmarkness being modelled as a context-dependent property a system
will:

￿ Learn about landmarkness in particular contexts, and store it.

￿ Produce views, which are triggered by certain thresholds of landmarkness ,

generating instances of a type landmark in a particular context.