Software Engineering Repositories
Marcia Heloísa Tavares de Figueredo Lima, Víctor M. Valle, Estevão Costa, Fylype Lira, Bruno Gadelha · 2019
Defining and classifying software requirements are critical tasks for determining software functionality and overall software architecture. In this sense, several types of research are being developed aiming to automate the classification of software requirements through the use of machine learning algorithms. However, the feasibility of such studies runs counter to the existence of a public database that is adequate in terms of quantity and quality of sample requirements. A requirement base widely used in this type of task is the PROMISE. However, the number of requirements is considered low for practical applications involving machine learning. This research presents an expansion of the PROMISE corpus. New software requirements were incorporated, and the resulting dataset was evaluated through the use of well-known machine learning algorithms. We observed some improvement in the performance of these algorithms regarding the identification of some types of software requirements.