A Domain Engineering Framework Based on Probabilistic Ontologies for Automated Selection of Features to Reuse

As the complexity demanded for new systems increases, techniques for reusing existing systems or artifacts become a key factor for achieving both productivity and quality. In such context, domain engineering (aka software product line engineering) is a discipline which focuses on reusing domain knowledge in order to quickly produce a family of systems, especially software-intensive systems. The nature of domain engineering involves developing conceptual models to capture vocabulary and meta-information about some particular domain and to define common and varying characteristics—or features—among systems. Since ontologies are, by definition, formal specifications of knowledge about some domain, they are a natural candidate for representing conceptual models in domain engineering. In this work we use probabilistic ontologies to represent features, requirements, meta-information about reusable software solutions, and relationships among all of them with respective degrees of uncertainty, in order to be able to use a combination of description logic and Bayesian reasoning for identifying a subset of reusable solutions (software artifacts) that best fits with an emerging problem’s specification. A proof of concept in the domain of Insider Threat Inference Enterprise Modeling is presented.