Eric Goldstein, Smadar Kedar, and Ray Bareiss
As CBR case libraries grow, they will become useful for an increasing number of tasks. An open issue in CBR, therefore, is how to index cases so that they can be retrieved for multiple tasks. Currently, most CBR systems index cases using indexing schemes which are specifically tailored to their one task. One way current CBR systems can perform retrievals for multiple tasks is to use multiple task-specific indexing schemes. A key drawback to using multiple indexing schemes is the increased difficulty of the indexing task for the human indexer. We present a method which makes it easier for a human indexer to create multiple task-specific indices. In particular, we describe how the method can (1) leverage off one kind task-specific index to create another kind of taskspecific index, and (2) partially automate the creation of one kind of task-specific index. We will illustrate our method in the context of the Story Archive project at the Institute for the Learning Sciences.