A computational model for constructing novel associations
Kazjon S. Grace, Robert N. Saunders, John S. Gero · 2008
Abstract. This paper presents a computational model for the construc-tion of novel associations as a component of a larger project on analogy-making. Association-construction is driven by a reinterpretation-based model of subjective similarity. Associations are constructed by trans-forming the way the agent perceives the objects being associated so that they become similar. The paper describes a model where an agent devel-ops associations by an iterative process of attempting to relate objects in its environment. Associations are constructed by building appropriate transformative interpretations. Methods for the learning of transforma-tions are discussed. The capabilities and implications of the model are discussed through an example application to the domain of geometric proportional analogies. 1