Automatic requirements elicitation in agile processes

One of the generic phases of software engineering is the requirements analysis. This paper presents a new method for automatically retrieving functional requirements from the stakeholders using agile processes. The presented system is a machine learning system for the automation of some aspects of the software requirements phase in the software engineering process. This learning system encompasses knowledge acquisition and belief revision in a knowledge base. It is based on Tecuci's multistrategy task-adaptive learning by justification trees algorithm, known as Disciple-MTL, and supports a few of the practices that extreme programming (XP) requires. The aim of the algorithm is to collect information from the various stakeholders and integrate a variety of learning methods in the knowledge acquisition process, while involving certain and plausible reasoning. The result of the manipulation is a list of requirements essential to a software system.