Structure abstractions in retrieval of analogical software models

摘要

Within object-oriented software engineering, there may be a high potential for reuse of artifacts ranging from requirement specifications to code if one is able to identify analogies between systems at an early stage, i.e. in the analysis process. This paper addresses the problem of analogical retrieval of object-oriented OOram software analysis models from a repository of such. Effective retrieval of analogs is hard both for humans and computers, and humans as well as many computer models mainly use surface properties. However, cognitive science research on analogy shows that structural properties of a situation must be considered just as important. In this paper an algorithm for obtaining abstractions of the structural properties of an OOram model is presented. The algorithm returns high level descriptions of the communication patterns between objects in the analysis models. The descriptors used are chain, ring, star, and tree and in addition parameters. Experiments with synthetic data show that combining these abstractions with semantic information about the objects in the models improves retrieval.