Using Typicality Theory to Select the Best Match

This paper focuses on the problem of choosing the best match among a set of retrieved cases. The Select step is subtask of case retrieval that produces the case that suggests the solution for the input case. There are many different ways to accomplish this task and we propose an automatic means for it. Following the original motivation of paralleling the human similarity heuristic we argue that the selection of the best match is performed by humans choosing the solution that best represents the set of candidate solutions retrieved. The solution that best represent a given data set is the “most typical” solution. Therefore, we describe an application in a Case-Based Reasoning system using the Theory of Typicality to calculate the Most Typical Value of a given set to automatically perform the Select task. An example illustrates the application.