IMPROVING QUERY RESULTS FOR SKETCH-BASED. Malumbo Chipofya

Sketch-based spatial query systems provide an intuitive method of user interaction for
spatial databases. These systems must be capable of interpreting user sketches in a way
that matches the information that the user intended to provide. One challenge that must be
overcome is that humans always simplify the environments they have experienced and this
is reflected in the sketches they draw. One such simplification is manifested as aggregation
or combination of spatial objects into conceptually or spatially related groups.
In this thesis I develop a system that uses reasoning tools of the RCC-8 to evaluate sketchbased
queries and provide a method for minimizing the effects of aggregation by
determining whether a solution to a query can be expanded if some groups of regions are
assumed to be parts of a larger aggregate region. If such a group of regions is found, then
this group must be included in the solution. The solution is approximate because the
approach taken only verifies that assumed parts of an aggregate are not inconsistent with
the configuration of the whole solution. Only cases where the size of the solution equals the
size of the query minus one are analysed.
It is observed that correctly identifying aggregated regions leads to solutions that are more
similar to the original query sketch when the size of every other solution is smaller than the
size of the query or when a lower limit is placed on the acceptable size of a solution because
the new, expanded or refined solution becomes more complete with respect to the sketch
of the query.