Metin tabanlı test durumlarının kullanacağı yazılım modüllerinin tahmin edilmesi (Predicting the software modules to be exercised by test cases given in natural language)

This is the latest version of this item.

Official URL: http://ceur-ws.org/Vol-1980/YTM_2017_paper_9.pdf

Both static and dynamic program analysis techniques have been extensively used for improving software quality. Recently, other types of analysis approaches have also been developed to analyze the software artifacts written in natural language, such as requirements specification documents and user manuals. In this work, we present a hybrid approach, which combines text analysisbased approaches and traditional program analysis approaches to further improve software quality. In particular, we develop a natural language processingbased approach to analyze textual test cases created manually by practitioners and predict the modules that these test cases would exercise at runtime. Note that this information can then be used in a wide spectrum of quality assurance tasks, such prioritizing the test cases and predicting different types of structural coverage that the test cases would achieve. In the empirical studies conducted on a large and complex industrial software system, the proposed approach successfully predicted the modules to be used with an average F-measure of 0.741 and it was significantly better than random prediction in the statistical manner.