NFRfinder: a knowledge based strategy for mining non-functional requirements

A challenge in requirements elicitation is to identify quality requirements, i.e. non-functional requirements (hereafter, NFR). In general, stakeholders need NFRs, but these requirements are not always explicit, they can be part of the tacit knowledge. The usual strategy adopted by requirements engineers to elicit NFRs is to act proactively by asking stakeholders their interests in qualities based on lists or catalogs. NFRFinder is a semi-automated process strategy for mining keywords. The strategy uses the keywords to find possible NFRs in unstructured texts, e.g. the meeting minutes that occur during an elicitation task. The strategy relies on catalogs, according to the NFR Framework, as a supporting knowledge base. However, to gain more confidence on the NFRFinder, we have applied it to a set of structured texts. We report on the recall and precision of NFRFinder using a gold standard built from different actors, for requirements sentences. The results are promising, and we point out towards the evolution of NFRFinder.