Measuring Similarity Between Software Artifacts

This paper presents a model of estimating the similarity of software artifacts so as to promote their analogical reuse. The model permits comparisons between artifacts developed at the various stages of the software development (i.e specifications, designs and code) from conceptual descriptions of these artifacts. This is achieved by using metrics measuring the distance between such descriptions with respect to general conceptual modeling abstractions (i.e the classification, the generalization and the attribution) underlying them[21]. Similarity estimates are influenced by measures of salience of the involved attributes. Salience is measured as belief on three domain independent properties of attributes (i.e the charactericity, the abstractness and the causality), suggested as predictive of their significance[22]. A prototype of the model is presented together with an example of using similarity to support the specification of requirements by reuse