H. Donfack, R. Nkambou, and V. Bevo (Canada)
CBR, Case retrieval, Ontology, Software engineering, dynamic inductive tree, Inverted list
The quality and performance of a CBR1 system depend largely on its capability to efficiently find interesting solutions (similar cases) for solving new problems. Thus, the search algorithm implemented in a CBR tool is an essential part of the system. The existing algorithms rely on characteristics (static or induced) of the problem to find out similar cases. In software engineering, and especially for software engineering experience capitalization considering the whole software development lifecycle, we think that though these characteristics (features) are important, they are not sufficient to retrieve efficiently the good case. We introduce in this paper a promising approach to address this issue. Our approach is based on dynamic inductive tree and inverted list. It is implemented in a two steps process: a dynamic inductive tree is used for the first step while an inverted list is used for the second step. The first step is based on problem characteristics categorized in subsets according to Cunningham's2 classification (descriptive aspect). In the second step, concepts of the considered domain ontology (structural aspect) are used to filter the results of the first step
Important Links:
Go Back
