A framework for constructing semantically composable feature models from natural language requirements

Abstract

Software Product Line Engineering (SPLE) requires the construction of feature models from large, unstructured and heterogeneous documents, and the reliable derivation of product variants from the resulting model. This can be an arduous task when performed manually, and can be error-prone in the presence of a change in requirements. In this paper we introduce a tool suite which automatically processes natural-language requirements documents into a candidate feature model, which can be refined by the requirements engineer. The framework also guides the process of identifying variant concerns and their composition with other features. We also provide language support for specifying semantic variant feature compositions which are resilient to change. We show that feature models produced by this framework compare favourably with those produced by domain experts by application to a real-life industrial example.