Surveying the Opportunities for Textual CBR: A Position Paper
Robin Burke · 1998
Riesbeck (1997) writes, that the.goal for AI and cBR in particular should be understood as "building components that reduce.the stupidity of the systems in which they ’ function. " (pg. 377) This is an excellent description of the niche that textual ease-based reasoning can and should occupy.. This position paper explores some of directions in which this goal can be pursued, with special attention to the role of task orientation. Progress in information retrieval and databases have brought us power-ful, flexible tools for managing large amounts of textual data, yet from a user’s point of view, these systems often seem stupid. They aim for a least-common-denominator interface, ensuring that every possible query can be posed, but do little io help satisfy user’s most common goals, They treat data uniformly, putting the burden on the user to distinguish structural ele-ments that are relevant for a particular task. As a result, users must master complex querying conventions to accomplish even simple tasks. To this problem, case-based reasoning brings the legacy of planning and problem-solving research. Case-based reasoners, like human reasoners, seek retrieval because they have a problem to solve. CBR systems build their indexing and retrieval mechanisms within knowledge frameworks that iden-tify important features and problem-solving strategies. An effective textual 13