Arborcraft: Automatic Feature Models From Textual Requirements Documents

One of the tasks facing requirements engineers working in the field of Product Line Engineering is the creation of feature models, which represent the domain of the product line and from which product configurations can be derived. Requirements documents, which are to be mined for this information, are often very large and written in potentially ambiguous natural language, and can be written over a long period of time by various authors. This makes the engineer's task very arduous, and a clear separation of concerns is often difficult to infer from textual documents. We present ArborCraft, a tool which automatically mines textual requirements documents, identifies features based on natural-language processing techniques, and produces a candidate feature model. We show that this can significantly reduce the burden on the requirements engineer and promote separation of concerns early in the Product Line Engineering process.