9 Modeling analogy as probabilistic grammar
Adam Albright · 2009
Abstract Models of analogy must be non-deterministic enough to handle gradient data, but must also explain why analogy obeys some striking restrictions: only a tiny subset of logically possible analogies are actually attested. This chapter discusses several unattested types of analogy, and considers their implications for formal models. Gradience and notable restrictions are best modeled using a grammar of probabilistic rules.