Geography Reference
In-Depth Information
if the domain is restricted, thus, domain-specific systems have an advantage. But
beyond domain-specific symbology, there are graphical conventions and standard
visual symbols that help understanding a sketch. For example, stick figures typically
represent humans, and sequences of snapshots often depict dynamics [ 28 ] .
To illustrate let us look at the CogSketch system [ 29 ] . It is a sketch understanding
system with an in principle open domain. For conceptual understanding, it relies
on the user to label drawn objects. Users have to select a fitting concept from
CogSketch's knowledge base, which is derived from OpenCyc 11 and contains
over 58,000 concepts. Three different interface techniques have been explored: (1)
concept lists providing access to a (small) subset of the knowledge base (selected for
a particular application); (2) direct access letting a user type in a concept name and
the system providing string completion; (3) glyph bars providing pictorial labels for
knowledge base concepts. The latter is similar to the first interface mode, however,
users do not need to know anything about the knowledge base, only to understand
the glyphs.
With visual and conceptual understanding, it becomes possible for a service to
interpret a user's sketched input and to interact by providing reasonable answers.
The original query-by-sketch application by Egenhofer [ 23 ] translated the concep-
tual understanding of the sketch (represented as a semantic network in their case)
into SQL (or SQL-like statements) for database lookups. Sketching here was seen
as an alternative, more user-friendly interaction mode for Geographic Information
Systems.
Later approaches suggested directly matching a sketch against other spatial
representations, for example, a base map [ 45 , 65 ] . In particular, this may be
used for localization purposes, either of particular places or, indeed, the users
themselves [ 46 ] . Landmarks are a common element in people's sketch maps [ 50 , 63 ] .
With sketches drawn for localization, landmarks are even more likely to be crucial
parts of the sketch because next to the configuration of the street network these
are the geographic objects most likely to be remembered or perceived by the users.
Given that landmarks are salient objects, they are also ideal candidates to restrict
the search space. Thus, the ability to understand graphical landmark references is
an important feature of spatial services employing a sketch-based interface.
6.4
Evaluating Landmark-Based Human-Computer
Interaction
Throughout this topic we have argued that incorporating landmarks is a necessity
for truly intelligent spatial services. We have also argued that services producing
and understanding landmarks deliver significant benefits for the human partner in
human-computer interaction. In this section, we will provide some evidence for
11 http://www.cyc.com/platform/opencyc , last visited 8/1/2014.
 
 
 
Search WWH ::




Custom Search