An empirical study of automated classification tools for informal requirements in large scale systems

Abstract

An important issue in software development process is to determine user and business requirements. This involves the basic tasks of elicitation, analysis and validation and documentation that determine the goals, functionalities and constraints of the software system. The traditional requirements elicitation strategies and techniques, if manually executed, work well in small or medium projects but are ineffective in large projects. Several works have been proposed to scale-up requirements of elicitation methods so as to facilitate the inclusion of many thousands of relevant stakeholders, who may be remotely distributed across the globe in the authoring of requirements. Unfortunately, such an open requirements elicitation method has been found to generate the informal requirements requests faster than can meaningfully be analysed, thereby defeating its purpose. To facilitate effective analysis, several methods have been proposed for decomposing the overwhelming requirements requests arising from the open and inclusive elicitation methods. This paper therefore carries out an experimental investigation of these methods so as to authenticate the strengths and weaknesses of each and to know where they need further support.