Analogy as Functional Recategorization: Abstraction with HowNet Semantics

R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 326 – 333, 2005. © Springer-Verlag Berlin Heidelberg 2005 Analogy as Functional Recategorization: Abstraction with HowNet Semantics Tony Veale Department of Computer Science, University College Dublin, Belfield, D4, Dublin, Ireland Tony.Veale@UCD.ie http://www.cs.ucd.ie/staff/tveale.html Abstract. One generally accepted hallmark of creative thinking is an ability to look beyond conventional labels and recategorize a concept based on its behav- iour and functional potential. So while taxonomies are useful in any domain of reasoning, they typically represent the conventional label set that creative think- ing attempts to look beyond. So if a linguistic taxonomy like WordNet [1] is to be useful in driving linguistic creativity, it must support some basis for recate- gorization, to allow an agent to reorganize its category structures in a way that unlocks the functional potential of objects, or that recognizes similarity between literally dissimilar ideas. In this paper we consider how recategorization can be used to generate analogies using the HowNet [2] ontology, a lexical resource like WordNet that in addition to being bilingual (Chinese/English) also provides explicit semantic definitions for each of the terms that it defines. 1 Introduction Analogy is a knowledge-hungry process that exploits a conceptual system’s ability to perform controlled generalization in one domain and re-specialization into another. The result is a taxonomic leap within an ontology that transfers semantic content from one term onto another. While all taxonomies allow vertical movement, a system must fully understand the effects of generalization on a given concept before any analogy or metaphor can be considered either deliberate or meaningful. So to properly support analogy, a taxonomy must provide a basis of abstracting not just to conventional cate- gories, like Person, Animal or Tool, but to categories representing the specific causal behaviour of concepts such as think-agent, pain-experiencer, cutting-instrument, and so on. Thus, a surgeon can be meaningfully described as a repairman since both occu- pations have the function of restoring an object to an earlier and better state; a foot- baller can be meaningfully described as a gladiator or a warrior since each exhibits competitive behaviour; and a scalpel can be compared to a sabre, a sword or a cleaver since each has a cutting behaviour; and so on. Theories of metaphor and analogy are typically based either on structure-mapping [3,4] or on abstraction e.g., [5,6,7,8,9,10]). While the former is most associated with analogy, the latter has been a near-constant in the computational treatment of meta- phor. Structure-mapping assumes that the causal behaviour of a concept is expressed in an explicit, graph-theoretic form so that unifying sub-graph isomorphisms can be Analogy as Functional Recategorization: Abstraction with HowNet Semantics 327 found between different representations. In contrast, abstraction theories assume that analogous concepts, even when far removed in ontological terms, will nonetheless share a common hypernym that captures their causal similarity. Thus, we should ex- pect an analogous pairing like surgeon and butcher to have different immediate hy- pernyms but to ultimately share an abstraction like cutting-agent (see [8,9]). …

Content truncated to 500 words