Evolutionary requirements elicitation and analysis based on crowdsourced user feedback: An approach and tool support

In recent years, the software market has experienced rapid expansion, leading to an increasing number of both end users and the feedback they generate. This provides researchers and practitioners in the requirements engineering (RE) community with a rich source of data to identify and understand user requirements. This dissertation addresses a critical challenge in an emerging and promising trend within the requirements engineering (RE) community — crowdsourced RE for mobile apps, one of the fastest-growing software types: how to effectively leverage the vast volume of user feedback to support RE activities, particularly for mobile apps.

Unlike traditional methods for RE activities, which are typically performed manually by human RE experts, automated methods are essential in Crowdsourced RE for mobile apps to effectively facilitate RE activities, particularly the rapid extraction and analysis of requirements from the vast volume of user feedback, enabling timely responses for software maintenance and evolution, especially app updates. Moreover, it is also valuable to investigate whether these requirements derived from user feedback are actually considered and, ideally, implemented in the next iterations of mobile apps. While several prior studies have reported the feasibility and effectiveness of automating requirements elicitation and analysis from app reviews, few have explored the use of app release notes to enhance automatic extraction and classification of requirements, particularly by distinguishing between functional requirements and different types of non-functional requirements. In addition, several studies have explored the links between app reviews and app updates, revealing valuable insights into whether and how app reviews influence the evolution and iterative development of mobile apps.

Motivated by this issue, this paper-based dissertation proposes an approach to enhance automated elicitation and analysis of requirements from crowdsourced user feedback for software maintenance and evolution. Through two systematic mapping studies, we first investigated the state-of-the-art of crowdsourced RE to identify which sources and metadata of crowdsourced user feedback, along with the most frequently employed techniques and tools for processing and analyzing the user feedback. Based on these findings, this dissertation proposed an approach to enhance the automatic elicitation of evolutionary requirements from app reviews, facilitate automatic classification of the extracted requirements into functional and non-functional requirement types, understand the characteristics of app updates, and finally identify the roles that app reviews serve in app updates from the developers’ perspectives. The approach was validated by a series of experiments on the constructed research dataset of app reviews and release notes to evaluate its performance and generalizability, and further identify the influence of app reviews on app updates.