160 likes | 324 Views
Knowledge-Intensive Case-Based Reasoning in CREEK. Knowledge-Intensive CBR in. Introduction. Assumes cases enriched with general domain knowledge Focus in these types of cases Study CREEK architecture and system (in general terms). Knowledge-Intensive CBR in.
E N D
Knowledge-Intensive Case-Based Reasoning in CREEK
Knowledge-Intensive CBR in Introduction • Assumes cases enriched with general domain knowledge • Focus in these types of cases • Study CREEK architecture and system (in general terms)
Knowledge-Intensive CBR in The Knowledge-Intensiveness dimension of CBR methods MBR IBL/IBR CREEK
Knowledge-Intensive CBR in What is knowledge in a CBR system? • Difficult to distinguish knowledge concept from information or data. • Attemps: • 1) Based on size or complexity (fail) • 2) “CREEK” option: • How and for What purpose the structures are used.
Knowledge-Intensive CBR in Knowledge Learning Information Data The Data-Information-Knowledge model Knowledge is learned information Elaboration Information is interpreted data Data Interpretation Data are syntactic entities whitout meaning for the system.
Knowledge-Intensive CBR in Case roles in CBR systems • Knowledge is needed for a system to reason. • (Interpreting data and deriving new info) • Roles of Cases: • 1) Cases as data • 2) Cases as information • 3) Cases as Knowledge • CB system architecture can facilitate a gradual transformation from a pure database or information system, to a full-fledged knowledge-based system.
Knowledge-Intensive CBR in Symbol Level Design and Implementation Knowledge Level Analysis and Modeling Sustained Learning Periodic Knowledge Revision EXPERIENCE COMPUTERINTERNALMODEL MENTAL MODEL CONCEPTUAL KNOWLEDGE MODEL Problem Solving New Case The Knowledge modeling cycle
Knowledge-Intensive CBR in The CREEK system • Is an architecture for knowledge-intesive case-based problem solving and learning, targeted at addressing problems in open weak-theory domains • It has different modules, the main ones are: • Object-level domain knowledge model • Strategy level model • Two reasoning meta-level models • It integrates problem solving and learning
Knowledge-Intensive CBR in Solved problems + traces New problem Problem Solving Sustained Learning MBR EBL CBR CBL General Domain Model Solved cases The CREEK system Diagnosis and Repair Strategy Combined Learning Combined Reasoning
Knowledge-Intensive CBR in The CREEK system • CBR tasks: • Retrieve • Reuse • Revise • Retain
Knowledge-Intensive CBR in The CREEK system • The case-based interpreter in CREEK contains three-step process: • Activating ( Activation Strength > Threshold ) • Generating and Explaining • Focusing
Knowledge-Intensive CBR in The CREEK system • Activate determines a relevant broad context for the problem • Explain will attempt to improve the match between the input case and the activated cases. • Focus selects the best case or rejects all of them
Knowledge-Intensive CBR in The CREEK system • Activation Strength is based on the number of matched relations and their relevance factor:
Knowledge-Intensive CBR in The CREEK system • The general domain knowledge is good enough to provide support to the case-base methods and also back-up if no similar case is found • Troll Creek allows running the case matching process at any time during system development
Knowledge-Intensive CBR in Recent and ongoing research • From Lisp implementation to Java • Revision of the knowledge representation and basic inference methods
Knowledge-Intensive CBR in Conclusion • Knowledge-intensive CBR is at the core of the CREEK framework, architecture, and system. • The current and future in CREEK: • - methods related to combined explanatory power of general domain knowledge and cases