Case - Based Reasoning - Survey and FutureDirectionsBrigitte Bartsch - Sp orl

This paper surveys the eld of case-based reasoning (CBR)-both in science and in industrial applications. It starts with a short introduction to the essential ideas and concepts CBR is built upon. Then follows a bit of history that is interesting for understanding the development and the current state of the eld. Its main part introduces and reviews the most important sub-elds of CBR: theoretical foundations, CBR for document retrieval, product selection, help-desk support, diagnosis , connguration, planning, and design. In the last part, we discuss why the eld has developed rather well and will have a promising future, particularly in new areas like self-service and e-commerce applications in the world wide web. 1 What is Case-Based Reasoning? Case-Based Reasoning (CBR) is an approach to develop knowledge-based systems that are able to retrieve and reuse solutions that have worked for similar situations in the past. This sounds-and is-simple but deserves some explanations of what are the prerequisites, the essential steps, and the diierences compared to other approaches for building knowledge-based systems. The rst and most important prerequisite is a collection of experiences, embodied in so-called cases, and stored in a case base. Every single case consists at least of a problem description part, called the problem, and a solution description part, called the solution. These two basic ingredients are usually enriched by an administrative part including, e.g., a case number, an explanation or justiication part that provides more information about the step from the problem to the solution, a context description part, and an evaluation part that contains information about the quality and reusability of the case. Moreover, there may be attachments in any format that multi-media computer equipment is able to record, view, and play. In addition to the case base, there may be some general knowledge in the form of models or rules or constraints available and used. The case base and the general knowledge constitute a partial domain model-with the consequence