Knowledge representation of geospatial entities is dependent on the ability to share their structural properties along with their functional properties, which define their usage for human-society. However, geospatial ontologies have mainly relied on taxonomy-based and mereology-based ontologies. While structural properties of entities such as shape, topography, and orientation are considered important tools for geospatial ontologies, existence of structural properties are not sufficient conditions for the existence of functional properties. Contrastingly, a parallel approach assumes independent existence of function-based concept hierarchies and builds on the premise that human activities associated to any given geospatial entity are essential for specification of the entity concept itself. This chapter compares two diverging approaches based on cases drawn from physical geography, transportation, and hydrology. The differences in core concepts and tools are discussed in relation to universal ontologies of geographic space. It is argued that function representation in geospatial ontologies, in combination with structure-based concepts of geospatial entities, is both necessary and challenging.
TopIntroduction
Concepts are assumed to have semantic properties and according to the classical theory, concepts are complex mental representations whose structure generally encodes a specification of necessary and sufficient conditions for their own application (Laurence & Margolis, 2003). Geospatial concepts are no exception and geospatial ontologies have been employed to specify concepts in the geospatial domain (Kuhn, 2005; Fonseca, Egenhofer, Agouris, & Câmara, 2002; Goodwin, 2005). Also, certain parallels can be drawn to the concepts represented in Geographic Information Systems (GIS) and Computer Aided Design (CAD) drawings. Both contain geometric representations of real world objects, but with different scales and applicability. The essential feature of both CAD and GIS data is that the semantics of the symbols is linked to the ontological commitment of the information system, thereby ensuring that linkage between the symbols (point, line, or polygon) and the real world object is well understood. Consequentially, for such data, ontologies can form the basis of (existence or non-existence of) semantic interoperability. Although, we will focus on the conceptual models in GIS rather than CAD, the connection between the two is important and shall be dealt briefly, later in this chapter.
Three major utilities of ontologies discussed by Uschold and Gruninger (1996) include (1) communication between people, (2) interoperability between computer systems, and (3) system engineering related. Knowledge representation about concepts has a similar foundation as provided by the theory of semantics in human communication. The fundamental ternary relationship between symbols, concepts and entities, explained by the semantic triangle (Ogden & Richards, 1946), has been argued to be critical for interoperability between information systems and information societies (Kuhn, 2005). Large-scale availability of geodata and geoprocessing services has lead to increasing focus on the need to ensure semantic interoperability (Brodaric, et al., 2007). This in turn provides the motivation to encode semantic knowledge in the form of ontologies. Agarwal (2005) has discussed the roles of ontologies in GI Science and their applicability in the geographical domain. Gruber (1993) defines ontologies as explicit specifications of conceptualizations. The crucial characteristic of GI is that it refers to entities on the Earth (and hence with a locational attribute). It is this referential meaning that needs to be made explicit and organized (Guarino, 1997) in geospatial ontologies. Nevertheless, we know that ontologies themselves are specified in different languages, logic framework, and using different specification techniques in order to solve the problem at hand. This decreases their universal applicability thus increasing impediments towards interoperability of geospatial data.
Much of the efforts in ontological engineering for the geospatial domain have been in the area of integration of geospatial databases Cruz, Sunna, and Chaudhry (2004), and more recently, geospatial web based services (geodata and geo-processing services) (Lutz, et al., 2006; Yue, et al., 2007). It follows that such specifications should be able to elucidate differences, if any, between concepts of two different systems and ontological reasoning should be able to prove if it is semantically consistent that data from one is used by another.
One key area governing the specification of ontologies is the type of knowledge that is being shared, with respect to the geospatial entities in question. While the underlying logical framework is closely related to this issue 1, in this chapter two different approaches are investigated, on the basis of a totally different paradigm. The concern is rather about the type of knowledge specified in ontologies, namely structural (based on unvarying characteristics such as height or colour of an entity or its parts) and functional (based on knowledge about functional relationships such as actions and utility afforded by the objects). The difference between the two approaches can be illustrated by two descriptions of motorways in the UK that are available.