6 Automated Geographical Information Fusion
111
The term “schema” is not identical to “ontology,” but the two terms are often
used near-interchangeably. A schema is a formally (or otherwise precisely) defined
taxonomy. Thus, a schema is an ontology in the sense of [32]. However, the term
“ontology” encompasses a broad spectrum of specification methods, from schemas
at one extreme through to general logical systems that can be used to define and
reason about sophisticated relationships and constraints between elements within a
taxonomy [47]. From this point onward, the term “ontology” is preferred because
this term covers both schemas and more sophisticated types of ontologies. However,
it should be noted that the ontologies in this chapter are simply schemas.
A critical step in the fusion process is to fuse the ontologies for the different
information sources. A variety of closely related terms are used in the literature to
refer to aspects of this task, including
• integration and alignment;
• merging and matching;
• transformation and mapping.
To further confuse the issue, almost all of these terms may appear in the literature
combined with any one of “ontology,” “schema,” or “semantic” (e.g. “ontology alignment,” “schema matching,” and “semantic integration”). The choice of which precise
terms are adopted by particular researchers is often more a matter of preference and
domain than a strict difference in definitions. However, a clear distinction is usually
made between the process of identifying the relationships between corresponding
elements in two heterogeneous ontologies (termed “alignment/matching/mapping”)
and the process of constructing a single combined ontology based on these identified relationships (termed “integration/merging/transformation”) [45, 51]. For consistency, in this chapter we use the terms “(ontology) integration” and “(ontology)
alignment” to distinguish these two concepts.
6.2.1 Ontology Integration and Mediators
The concept of a mediator, a software system that can assist humans in integrating
heterogeneous information sources, was first explicitly described by Gio Wiederhold [69]. Based on his general vision, dozens of different mediation systems have
been proposed and developed over the past decade (for a full survey of mediation systems see [66]). For example, TSIMMIS was one of the earliest mediator
systems to be researched. The core idea behind TSIMMIS was to mark up information sources with standardized tags, which included labels describing the semantics of each data item [12, 30]. Rather than using unstructured tagging to describe
information sources, subsequent mediators, such as SIMS [1], OBSERVER [48],
InfoSleuth [3], and OntoSeek [34], use predefined domain ontologies as the basis
for integration. More recently, a suite of Web-based languages and mediation technologies have emerged around the topic of the semantic Web (e.g. [31]). The primary
focus of all these systems is efficient integration of information across multiple information sources with heterogeneous ontologies. Ontology alignment is a prerequisite
for such systems to operate, but the question of how the alignment semantics are
111
The term “schema” is not identical to “ontology,” but the two terms are often
used near-interchangeably. A schema is a formally (or otherwise precisely) defined
taxonomy. Thus, a schema is an ontology in the sense of [32]. However, the term
“ontology” encompasses a broad spectrum of specification methods, from schemas
at one extreme through to general logical systems that can be used to define and
reason about sophisticated relationships and constraints between elements within a
taxonomy [47]. From this point onward, the term “ontology” is preferred because
this term covers both schemas and more sophisticated types of ontologies. However,
it should be noted that the ontologies in this chapter are simply schemas.
A critical step in the fusion process is to fuse the ontologies for the different
information sources. A variety of closely related terms are used in the literature to
refer to aspects of this task, including
• integration and alignment;
• merging and matching;
• transformation and mapping.
To further confuse the issue, almost all of these terms may appear in the literature
combined with any one of “ontology,” “schema,” or “semantic” (e.g. “ontology alignment,” “schema matching,” and “semantic integration”). The choice of which precise
terms are adopted by particular researchers is often more a matter of preference and
domain than a strict difference in definitions. However, a clear distinction is usually
made between the process of identifying the relationships between corresponding
elements in two heterogeneous ontologies (termed “alignment/matching/mapping”)
and the process of constructing a single combined ontology based on these identified relationships (termed “integration/merging/transformation”) [45, 51]. For consistency, in this chapter we use the terms “(ontology) integration” and “(ontology)
alignment” to distinguish these two concepts.
6.2.1 Ontology Integration and Mediators
The concept of a mediator, a software system that can assist humans in integrating
heterogeneous information sources, was first explicitly described by Gio Wiederhold [69]. Based on his general vision, dozens of different mediation systems have
been proposed and developed over the past decade (for a full survey of mediation systems see [66]). For example, TSIMMIS was one of the earliest mediator
systems to be researched. The core idea behind TSIMMIS was to mark up information sources with standardized tags, which included labels describing the semantics of each data item [12, 30]. Rather than using unstructured tagging to describe
information sources, subsequent mediators, such as SIMS [1], OBSERVER [48],
InfoSleuth [3], and OntoSeek [34], use predefined domain ontologies as the basis
for integration. More recently, a suite of Web-based languages and mediation technologies have emerged around the topic of the semantic Web (e.g. [31]). The primary
focus of all these systems is efficient integration of information across multiple information sources with heterogeneous ontologies. Ontology alignment is a prerequisite
for such systems to operate, but the question of how the alignment semantics are