Validating Feature Models With Respect to Textual Product Line Specifications

Feature models (FM) are a valuable resource in the analysis of software product lines (SPL). They provide a visual abstraction of the variation points in a family of related software products. FMs can be manually created by domain experts or extracted (semi-) automatically from textual documents such as product descriptions or requirements specifications. Nevertheless, there is no way to measure the accuracy of a FM with respect to the information described in the source documents. This paper proposes a method to quantify and visualize whether the elements in a FM (features and relationships) conform to the information available in a set of specification documents. Both the correctness (choice of representative elements) and completeness (no missing elements) of the FM are considered. Designers can use this feedback to fix defects in the FM or to detect incomplete or inconsistent information in the source documents.