Qualitative Geospatial Processing, Ontology and Spatial Metaphor
David Dodds, Open-Meta Computing Inc
There are many useful circumstances where a computer is able to wield competence about qualitative geospatial concepts and data. This is accomplished through the use of ontologies whose integrated usage coupled with qualitative computer processing (“non-crisp”) provides program capabilities reminiscent of some of the conceptual metaphorical usages of natural language.
The sciences of astronomy and physics have long used conceptual metaphor productively, in such terms as “event horizon”, “gravity well”, “curved space”, “time dilation”. The same technologies which provides the mechanisms for such term usage is available to geospatial processing. Productive conceptual metaphors can be used functionally on computers. This may be done by employing ontologies which provide requisite background knowledge to the system and qualitative logic processing. For example, a metaphorical comparison between two geospatial entities is qualitative and able to arrive at (useful) data when often strict-equality and simple scalar comparison either outright fails or returns a possibly nonsensical answer.
A code based example follows. It shows an XML RDFS element (IsNear) from the author's spatial ontology. Because it is XML it is compatible for processing with XML based GML material. Following this spatial ontology excerpt are function representations which show the meaning of (the function) “g1 (x)” which occurs in the comment of the IsNear predicate. The IsNear predicate (shown in the RDFS) has concomitant code (Java) which implements the calculation IsNear (as g1(x)) which returns a scalar value, not a boolean, which represents the degree of “nearness” found. The “degree” calculation is “situated”, that is, it is done within a context. The two geospatial objects involved have origins located at x1,y1; and x2,y2, respectively. These are formed into the centroidsdistance which informs the value of x in the function g1(x). Hence the computer is able to process the conceptual / visual metaphor of “nearness” (between object1 and 2) and instead of merely reporting near yes or near no, as would a strict equality / first order logic ontology system, it returns a scalar which can be processed within the context into a linguistic expression, such as “sort of near”, or “quite near”, or “not at all near”. The paper presentation explains in detail how all this is done.
</rdf:Property>
<rdf:Property ID="IsNear">
<rdfs:comment>has a degree of nearness (by value). g1(x)</rdfs:comment>
<rdfs:range rdf:resource="#www.open-meta.com/2001/IsNear/>
<rdfs:domain rdf:resource="#SvgEntity" />
</rdf:Property>
g1(centroidsdistance)
centroidsdistance = sqrt((x2-x1)^2+(y2-y1)^2)
IsNear = g1 = 1 - (1/2 + 1/PI*arctan((sqrt((x2-x1)^2+(y2-y1)^2))-k(1)/k(2)))
(IsNear(x1,y1,x2,y2))
The paper explains the computer processing involved to sensibly process the spatial input “A big dog is smaller than a small house.” and “A huge tree is smaller than a small mountain.” GML data is shown inexample of qualitative geospatial processing. Computer searches of GML data such as “Find all the smallish shallow water that is somewhat near to a gully or fairly small mountain.”may be of interest.