Supporting Dialogue Inferencing in Conversational Case-Based Reasoning

Dialogue inferencing is the knowledge-intensive process of inferring aspects of a user's problem from its partial description. Conversational case-based reasoning (CCBR) systems, which interactively and incrementally elicit a user's problem description , suuer from poor retrieval eeciency (i.e., they prompt the user with questions that the user has already implicitly answered) unless they perform dialogue inferencing. The standard method for dialogue inferencing in CCBR systems requires library designers to supply explicit inferencing rules. This approach is problematic (e.g., maintenance is diicult). We introduce an alternative approach in which the CCBR system guides the library designer in building a domain model. This model and the partial problem description are then given to a query retrieval system (PARKA-DB) to infer any implied answers during a conversation. In an initial empirical evaluation in the NaCoDAE CCBR tool, our approach improved retrieval eeciency without sacriicing retrieval precision. The distinguishing beneet of conversational case-based reasoning (CCBR) (Aha & Breslow, 1997) is that users are not required to initially provide a complete description of their problem. Instead, users enter text partially describing their problem and the system assists in further elaborating the problem during a conversation, which ends when the user selects a solution to apply. Several commercial CCBR tools exist; they have been particularly successful in help-desk and related applications (Watson, 1997). To behave eeciently, the CCBR tool should automatically infer problem description details from the user's text, and more generally infer these details whenever possible during a conversation. Otherwise, the user will be prompted to answer questions that they have already implicitly answered. A commercial solution to this dialogue inferencing problem requires the library designer to provide a complete and correct set of independent inferencing rules. But this approach complicates maintenance, and there are no guarantees concerning rule correctness or domain completeness. We introduce a model-based reasoning approach for solving this problem in which the library designer supplies a domain model of the case library and rules