Résumé
Analogical reasoning is considered as one of the main mechanisms underlying human intelligence and creativity, allowing the paradigm shift essential to a creative process. More specific is the notion of analogical proportion like “2 is to 4 as 5 is to 10” or “read is to reader as lecture is to lecturer”: such statements can be precisely described within an algebraic framework. When the proportion holds between concepts as in “engine is to car as heart is to human” or “wine is to France as beer is to England”, applying an algebraic framework is less straightforward and a new way to understand analogical proportions on the basis of Kolmogorov complexity theory may seem more appropriate. This viewpoint has been used to develop a classifier detecting analogies in natural language. Despite their apparent difference, it is quite clear that the two viewpoints should be strongly related. In this paper, we investigate the link between a purely abstract view of analogical proportions and a definition based on Kolmogorov complexity theory. This theory is used as a backbone to experiment a classifier of natural language analogies whose results are consistent with the abstract setting.