Information Technology Reference
In-Depth Information
result of some automatic blending process, but was intentionally developed by an
ontology engineer. In case that the ontology engineer intended to develop an ontology
of houseboats
B hb , this would be considered successful. However, if the engineer
intended to develop an ontology of boathouses, then
B hb would be a poor outcome.
In other words, the ontology consisting of
B hb and the intention houseboat would
have high fidelity, but the ontology consisting of
B hb and the intention boathouse
would have low fidelity. Thus, the evaluation of the theory
B hb is dependent on the
domain it is supposed to represent.
The lesson from this thought experiment is that the evaluation of the results of
ontology blending is dependent on the intended goal and, more generally, on the
requirements that one expects the outcome of the blending process to meet. One
way to capture these requirement is similar to competency questions, which are
widely used in ontology engineering [ 26 ]. Competency questions are usually initially
captured in natural language; they specify examples for questions that an ontology
needs to be able to answer in a given scenario. By formalising the competency
questions one can use automatic theorem provers to evaluate whether the ontology
meets the intended interpretation.
The requirements that are used to select between the different blends fall, roughly,
into two categories: ontological constraints and consequence requirements . Onto-
logical constraints prevent the blends from becoming 'too creative' by narrowing
the space for conceptual blending. E.g., it may be desirable to ensure that the
is_inhabited_by relationship is asymmetric and that is_navigated_by
is irreflexive. To achieve that any blendoid can be checked for logical consistency
with the following ontology:
ontology OntologicalConstraints =
ObjectProperty :is_inhabited_by
Characteristics : Asymmetric
ObjectProperty :is_navigated_by
Characteristics : Irreflexive
Given these requirements, any blendoid that involves a house that lives in itself, or
any boat navigated by itself would be discarded.
Consequence requirements specify the kind of characteristics the blendoid is sup-
posed to have. E.g., assume the purpose of the conceptual blending is to find alter-
native housing arrangements, because high land prices make newly built houses
unaffordable. In this case, the requirement could be 'a residence that is not located
on a plot of land', which can be expressed in OWL as follows:
ontology ConsequenceRequirements =
[...]
Class PlotFreeResidence
EquivalentTo :Residence
and (is_located_on only ( not (Plot )))
For the evaluation of a blendoid against requirements (both ontological constraints
and consequence requirements), it is often not sufficient to just consider the infor-
mation that is contained in the blendoid itself. Some background knowledge usually
needs to be added in order to evaluate a blendoid.
Search WWH ::




Custom Search