Eliciting requirements from Stakeholders’ responses using natural language processing

Abstract

Most software systems have different stakeholders with a variety of concerns. The process of collecting requirements from a large number of stakeholders is vital but challenging. We propose an efficient, automatic approach to collecting requirements from different stakeholders’ responses to a specific question. We use natural language processing techniques to get the stakeholder response that represents most other stakeholders’ responses. This study improves existing practices in three ways: Firstly, it reduces the human effort needed to collect the requirements; secondly, it reduces the time required to carry out this task with a large number of stakeholders; thirdly, it underlines the importance of using of data mining techniques in various software engineering steps. Our approach uses tokenization, stop word removal, and word lemmatization to create a list of frequently accruing words. It then creates a similarity matrix to calculate the score value for each response and selects the answer with the highest score. Our experiments show that using this approach significantly reduces the time and effort needed to collect requirements and does so with a sufficient degree of accuracy.