Information Technology Reference
In-Depth Information
obtained by the rule that given any element x of an involved set, any images of x
under the involved functions are identified. The quotient is computed by selecting a
representative of each equivalence class.
A difficulty that arises is that we have to make a choice of these representatives,
and therefore of names for the symbols in the colimit, since a symbol is often not
identically mapped in the base diagram of the blendoid. The convention in
is
that, in case of ambiguity, from among all symbols of the equivalence class, that
name of the symbol is chosen which is the most frequently occurring one. In any
case, the user has control over the namespace because the symbols in the colimit can
later be renamed. We can see this for our boathouse example above, where Agent
appears most often in the diagram and therefore the symbol has been correspond-
ingly renamed.
Hets
9.5.2 Evaluating the Blending Space
Optimality principles (see Sect. 9.3.2 ), in particular structural ones, can be used to
rank candidate blendoids on-the-fly during the ontology blending process. However,
even if they improve on existing logical and heuristic methods, optimality principles
will only narrow down the potential candidates and not tell us whether the result is
a 'successful' blend of the ontologies. For example, assume that we had optimality
principles that would show that from the roughly 1,000 candidate blendoids of House
and Boat that Goguen computed, only two candidates
B hb and
B bh are optimal. Is
either
B bh any good? And, if so, which of them should we use? To answer
these question, it seems natural to apply ontology evaluation techniques.
Ontologies are human-intelligible and machine-interpretable representations of
some portions and aspects of a domain that are used as part of information systems.
To be more specific, an ontology is a logical theory written in some knowledge
representation language, which is associated with some intended interpretation. The
intended interpretation is partially captured in the choice of symbols and natural
language text (often in the form of annotations or comments). The evaluation of an
ontology covers both the logical theory and the intended interpretation, their rela-
tionship to each other, and how they relate to the requirements that are derived from
the intended use within a given information system. Therefore, ontology evaluation
is concerned not only with formal properties of logical theories (e.g., logical con-
sistency), but, among other aspects, with the fidelity of an ontology; that is whether
the formal theory accurately represents the intended domain [ 56 ]. For example, if
B hb is an excellent representation of the concept houseboat , then
B hb or
B hb provides a
poor representation of the concept boathouse . Thus, any evaluation of the blend
B hb
depends on what domain
B hb is intended to represent.
Given these considerations,
B bh are not ontologies, they are logical
theories that are the result of the blending of two logical theories that are part of
ontologies. This is illustrated by the following thought-experiment: let's assume
the theory
B hb and
B hb captures the concept houseboat very well, and that
B hb is not the
Search WWH ::




Custom Search