KITE CBR Instruction

The KITE CBR engine is designed to emulate human remembering that people recall similar problems they have solved in the past and adapt the old solutions to solve the new problem. The goal of the KITE Super Search is to retrieve cases based on semantic meanings (similarities) of cases. As illustrated in the figure, there are four major components in the KITE CBR engine - 1) case library, 2) feature vector space, 3) user interface, and 4) search engine.
First, all cases indexed in the case library are converted into numeric values called case feature vectors according to distance matrixes of the case index structure and put in the high dimensional vector space. Second, a user querying the case library will use the interface (Super Search) to identify the aspects of the technology integration problem (context or situation) that are most relevant. Then, the user interface turns the problem into a query case which in turn is converted into a query feature vector. Third, the query vector is matched against all case vectors in the high dimensional vector space using the nearest neighbor algorithm that calculates the weighted distances between the query vector and all case vectors. Fourth, the search engine returns all matched cases with case numbers and abstracts ranked in distance. The shorter distance (higher percentage) means closer match. Then, the user can read abstracts of the retrieved cases and choose a matched case number to open the solution case.