Concept similarity measures the understanding between two agents

When knowledge in each agent is represented by an ontology of concepts and relations, concept communication can not be fulfilled through exchanging concepts (ontology nodes). Instead, agents try to communicate with each other through a common language, which is often ambiguous (such as a natural language), to share knowledge. This ambiguous language, and the different concepts they master, give rise to imperfect understanding among them: How well concepts in ontology O-A map(1) to which of O-B? Using a method sim that finds the most similar concept in O-B, corresponding to another concept in O-A, we present two algorithms, one to measure the similarity between both concepts; another to gauge du, the degree of understanding that agent A has about B's ontology. The procedures use word comparison, since no agent can measure du directly. Method sim is also compared with conf, a method that finds the confusion among words in a hierarchy. Examples follow.