The ability of analogy making is considered to be an essential part of intelligent behavior. By mapping concepts and knowledge about a well-known domain into a target domain, new hypotheses about that domain can be inferred. Such a transfer may result in a new conceptualization of the target domain and therefore can be seen as a source of creativity. Moreover, the discovery of common structures can initiate a generalization process and support the introduction of abstract concepts, which are not directly observable. While the creation of an analogical mapping is well examined, provided the corresponding representations of both domains are already chosen in a way that the structural compatibility is obvious, such situations seem to be somewhat artificial. In fact, the structural commonalities characterizing two analogous domains are usually not obvious in advance, but become visible as a result of the analogy making process. The conceptualization (i.e. the representation) must be modified and adapted to make implicit analogous structures explicit. It is argued that an essential part of establishing an analogy is a change of representation of one or both domains to allow for discovering the common structure (Indurkhya 1992). In this paper we sketch an approach to deal with the problem of re-representation in a logic-based framework for analogy making that applies anti-unification to compute the analogical relation.
[1]
Robert W. Hasker.
The replay of program derivations
,
1995
.
[2]
J. van Leeuwen,et al.
Theoretical Computer Science
,
2003,
Lecture Notes in Computer Science.
[3]
Gordon Plotkin,et al.
A Note on Inductive Generalization
,
2008
.
[4]
Kai-Uwe Kühnberger,et al.
Metaphors and heuristic-driven theory projection (HDTP)
,
2006,
Theor. Comput. Sci..
[5]
Jochen Burghardt,et al.
E-generalization using grammars
,
2005,
Artif. Intell..
[6]
B. Indurkhya,et al.
Metaphor and Cognition: An Interactionist Approach
,
1993,
CL.