Case-based reasoning in software component retrieval

Abstract

Software Component Retrieval refers to identifying the appropriate software components in a library which is plays a critical role in the success of the final system. Several methods have been proposed for retrieval and selection of components. The success factor of these techniques is mainly whether they can retrieve components that meet the functionalities required by the user of the system, With the advancement of these search and retrieval techniques, it has been mostly possible to retrieve components that meet the user queries. But it is argued in this research that current component retrieval techniques have two short falls which are "inability to provide decision support" and "inability to explain retrieval". The first problem refers to that although the retrieval system retrieves components but it does not provide any basis for decision support. The reason behind this problem is that the current retrieval methods do not keep track of the previous decision information of the final selection. The second problem refers to that the existing retrieval techniques do not provide any explanation of their retrieval which causes the user of the system to understand to what extent his query is related to the retrieval. In this research, to solve the problem of "inability to provide decision support", Case-based Reasoning component retrieval model is proposed that reuses previous component selection decision while retrieving components to provide a basis for decision and to help the developer deciding for selecting the most appropriate component among the retrieved ones. To solve the problem of "inability to explain retrieval", the reasoner of the proposed model uses domain ontology for semantic similarity calculation between the query and case, and at the same time generates explanation of the retrieval. For evaluation purpose, a prototype is developed and the proposed model is compared with existing approaches.