Integrating IR and CBR to locate relevant texts and passages

The paper presents the SPIRE system, a hybrid case based reasoning (CBR) and information retrieval (IR) system that: (1) from a large text collection, retrieves documents that are relevant to a presented problem case; and (2) highlights within those retrieved documents passages that contain relevant information about specific case features. We present an overview of SPIRE, run through an extended example, and present results comparing SPIRE's with human performance. We also compare the results obtained by varying the method by which queries are generated. SPIRE aids not only problem solving but knowledge acquisition by focusing a text extractor-person or program-on areas of text where needed information is likely to be found. Once extracted, this information can be used to create new cases or database objects, thus closing the loop in the problem solving knowledge acquisition process.