Feature and variability extraction from natural language requirements

Abstract

Requirements or requirement specifications are usually the very first artifact that is created when starting a software project. It results from intensive discussions with customers about all facets of the software system to be developed, mostly in form of plain text. Hence, if properly done, requirement specifications reflect on all features of a software system as well as how these features depend on each other. Consequently, requirements are a valuable source for extracting variability to support systematic reuse in variant-rich software systems. We research on techniques that employ syntactical as well as semantical information from such textual requirements in order to (1) extract features and map requirements to them, and (2) extract variability to understand how features are related to each other, and thus, which variants can be possibly created.