Empirical Principles and an Industrial Case Study in Retrieving Equivalent Requirements via Natural Language Processing Techniques

摘要: Though very important in software engineering, linking artifacts of the same type (clone detection) or different types (traceability recovery) is extremely tedious, error-prone, and effort-intensive. Past research focused on supporting analysts with techniques based Natural Language Processing (NLP) to identify candidate links. Because many NLP exist their performance varies according context, it crucial define use reliable evaluation procedures. The aim this paper propose a set seven principles for evaluating identifying equivalent requirements. In paper, we conjecture, verify, that perform given dataset both ability odds requirements correctly. For instance, when are high, then reasonable expect will result good performance. Our key idea measure random factor specific dataset(s) adjust observed accordingly. To support application report practical case study evaluates large number context an Italian company defense aerospace domain. current However, most proposed seem applicable any estimation technique aimed at binary decision (e.g., equivalent/nonequivalent), estimate range [0,1] similarity provided by NLP), used as benchmark (i.e., testbed), independently estimator text) method NLP).