Exploiting similarity metrics and case-bases for knowledge sharing between case-based reasoners

摘要

Differences in naming, structure, precision, and representation, collectively referred to as semantic heterogeneity, have long hindered efforts to share knowledge between reasoning agents. Most current techniques require the construction of an expensive, and highly formalized interlingua before any communication can have meaning between semantically heterogeneous agents. We present here a method for case-based reasoners to share knowledge without the need for a prior interlingua. Using the similarity metrics and the cases known to each agent, we demonstrate how classes in one knowledge base can be mapped into classes of another knowledge base with the help of a critic function. In the case that the critic function is mechanically realizable (e.g. a high fidelity simulation of the domain of interest), our method becomes well-suited for highly-autonomous case-based reasoners.