Conceptual Aspects: How Landmarks Can Be Described in Data Models - Landmarks: GIScience for Intelligent Services - page 119

Chapter 5 will be dedicated to compute the values of landmarkness in this
model. In principle multiple approaches are possible: studying the properties of
objects that can be sensed or learned, category-based classifications, studying the
graphic or verbal expressions of people searching for relata, or direct surveying
via questionnaires or user-contributed content. Each of these methods applies to a
particular context—e.g., a survey among pedestrians will reveal another list as a
survey among motorists, and the context can now be stored and accessed together
with the landmarkness value. This means if the goal is collecting the landmarks of an
individual's mental representation (which, as we have seen in the previous section,
is impossible in its entirety) a spatial database can manage and will not confuse
these landmarks in conversations with other persons. The following chapter will
especially expand this model from a single measure of landmarkness to a vector.
Such an extension is relatively straight forward and does not change the principles
of the model:
data Lns = Lns Double Double Double
- new data type: landmarkness captured by a three-valued vector,
- for example, by visual, semantic and structural salience (Ch. 5 )
The extension does not change the principles of the model, only the data string
in entities gets longer:
Entity ... [Ls Pedestrian (Lns 0.7 0.8 0.4) ...,
Ls Motorist (Lns 0.9 0.8 0.6 ...]
Accordingly, some operations such as hasLns need to be adapted as well to cope
with the extended data type. Semantically hasLns can stay the same. If one of the
landmarkness values is larger than 0 it would return True . New operations should
cater for such a new data type, like computing an overall landmarkness value from
a vector of values. The next section will do this by suggesting weighted averages.
Then, hasLns could be redefined to return True if the overall landmarkness value is
larger than 0.
Chapter 6 , finally, will be dedicated to use this model in intelligent spatial
communication with persons. Most of this work will be about the selection of
appropriate landmark references in machine-generated messages. Here all three
data types—the landmarkness context, the landmarkness value, and the preferred
reference—will be engaged with. Further selection criteria will be generated by
the request itself, which is a part of the communication context that can hardly
be anticipated and pre-computed. For example, if someone asks for a route to the