- 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。
~ DAVID W. AHAz , HECTOR MUNOZ-AVILAyz z Navy Center for Applied Research in Arti cial Intelligence, Naval Research Laboratory, Code 5510, 4555 Overlook Ave, SW, Washington, DC 20375-5337 aha@aic.nrl.navy.mil yz Department of Computer Science, University of Maryland, College Park MD 20742-3255 munoz@cs.umd.edu Case-based reasoning (CBR) is an AI problemsolving paradigm that stresses the reuse of stored cases, comprising of hproblem,solutioni pairs, to solve new, similar problems. Important technical issues related to this subject include representation, indexing, retrieval, revision, and retention. Currently, the most important international CBR meetings include the international CBR conferences and the European CBR Workshops, and workshops have also been held in several countries (e.g., the United States, Germany, United Kingdom, Italy). This special issue addresses interactive CBR research, which we de ne as an extension of the CBR paradigm in which a user is actively involved with the inferencing process. Interest in interactive CBR has recently increased, in large part due to commercial motivation; the most commercially successful application of CBR tools has targeted the customer support market niche (Watson, 1997). This has been pursued vigorously by Inference Corporation and, more recently, several other companies that market CBR shells. Help desk systems are typically interactive; they require interaction between customers and, for example, call center personnel. Although the CBR techniques in these systems have been historically simple from a researcher's perspective, this does not imply a lack of interesting applied research issues. In particular, interactive CBR tools must address several topics not addressed by non-interactive tools (e.g., dialogue management, user modeling) and, due to their applied nature, must address integration issues with additional systems. For example, perhaps the most popular type of interactive CBR systems are what we refer to as conversational CBR (CCBR) systems, which can be characterized as interactive systems that, via a mixed-initiative dialogue, guide users through a question-answering sequence in a case retrieval context. Although Aha and Breslow (1997) helped to popularize the phrase \CCBR", their inspiration was from Inference Corporation, and other groups had been previously researching this topic (e.g., Shimazu et al., 1994) or began studying it at approximately the same time (e.g., Racine & Yang, 1997). Four papers in this special issue relate to CCBR. We include ours rst because it summarizes research on simple CCBR tools, focusing on contributions for simplifying the case authoring process, enhancing human-machine conversations, and extending CCBR to address decision support tasks. In the next article, Shimazu, Shibata, and Nihei describe recent advances in ExpertGuide, a more advanced CCBR tool, focusing on demonstrating how CCBR tools can be used to develop WWW mentoring systems in a knowledge management context. They describe multilink retrieval capabilities to allow libraries to be
c
Fra Baidu bibliotek
Applied Intelligence, ??, 1{2 (1999) 1999 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Applied Intelligence: Special Issue on Interactive Case-Based Reasoning
2
Special Issue on Interactive Case-Based Reasoning
searched from multiple viewpoints, entropy algorithms for ranking questions, and indexing cases using scripts. Next, Yang and Wu address problems with very large case bases, and describe advances to CaseAdvisor. They introduce a real time algorithm for creating a decision forest to cluster cases, and then use an information gain approach to select questions. These articles all include empirical evaluations that demonstrate the utilities of the algorithms described. McSherry's article di ers from the others; it analyzes needs for CCBR systems in the context of sequential diagnosis tasks and describes how relevant advances in rule-based expert systems can be used to improve CCBR behavior. His focus is CBR-Strategist, which embodies these advances. The remaining articles focus on interactive CBR is a broader context. For example, Leake and Wilson's DRAMA system is exciting for its introduction of concept maps to the CBR literature, and in demonstrating how they can be used to support interactive retrieval and adaptation (i.e., in the context of aerospace design tasks). McKenna and Smyth then extend their well-known research on case competence modeling by demonstrating how a visualization tool, in their CASCADE program, can provide valuable feedback to users during the case authoring process. We believe this pioneering work will inspire several others to investigate how other visualization techniques can synergize with interactive CBR tools. The nal article describes the latest developments of SaxEx, an impressive CBR tool for increasing the expressiveness of musical phrases. In their article, Arcos and Lopez de Mantaras describe how users can interactively parameterize the system, and report on the utility of this functionality. This rst special issue devoted to interactive CBR is highly appropriate for Applied Intelligence, given its application orientation. Researchers studying this subject have recognized