E Pluribus Unum - Computational Creativity Research: Towards Creative Machines - page 161

Distributed Ontology Language DOL can be used to declaratively specify blending
diagrams of various shapes, and discuss in detail how the workflow and creative act
of generating and evaluating a new, blended concept can be managed and computa-
tionally supported within Ontohub, a DOL -enabled theory repository with support
for a large number of logical languages and formal linking constructs.
9.1 Concept Invention via Blending
One broad area of phenomena that is often brought into connection with issues
of creativity and the emergence of new ideas concerns notions such as metaphor,
blending, category mistakes, similes, analogies and the like. In each of these,
seemingly inconsistent material is combined in a manner that results in a produc-
tive growth of information instead of simple logical contradiction. Approaches to
treat this phenomenon are varied but commonly come to the conclusion that more
or less well developed notions of 'structure' are crucial for bringing the growth
of information about—e.g., 'implication complexes' for metaphor [ 6 ], 'conceptual
spaces' 1 for blending [ 14 ], 'structure mapping' in analogy [ 20 ], and so on. On the
one hand, the less structure that is available, the less productive the combinations
appear to be; on the other, the presence of structure raises the challenge of how such
formal commitments can be productively 'overridden' or rearranged in order to avoid
contradiction.
In our ongoingwork on ontology and its formal underpinnings, we have been led to
a very similar set of questions. By 'ontology' we here refer to the now rather standard
notion of a formal specification of a shared understanding of the entities, relations
and general properties holding in some domain of interest (cf. [ 25 , 27 ]). Achieving
adequate treatments in various domains has demonstrated to us the need for hetero-
geneous ontological specifications that are capable of capturing distinct perspectives
on the phenomena being modelled. In an architectural context, for example, it is ben-
eficial to maintain distinct perspectives on structural integrity, spatial distribution,
movement patterns by the occupants of a building ('flow'), navigation networks (pos-
sibly varying according to 'normal' and 'emergency' conditions), 'visibility' patterns
(both for users and for sensors in the case of security) and many more [ 5 ]—each of
these perspectives can be modelled well by employing ontological engineering tech-
niques but there is no guarantee that they are simply compatible. Our work on natural
language dialogue systems involving spatial language comes to the same conclusion
[ 2 ], while similar concerns are already well known in Geographic Information Sci-
ence [ 18 , 37 ]. To support this fundamental 'multi-perspectivalism' we have therefore
been developing an entire toolset of more sophisticated combination methods [ 38 ],
leading to the formal definition of the notion of a 'hyperontology' in [ 43 ].
1 The usage of the term 'conceptual space' in blending theory is not to be confused with the usage
established by Gärdenfors [ 19 ].