Generating Requirements Out of Thin Air: Towards Automated Feature Identification for New Apps

TL;DR

This paper explores using app store data and crowd insights to automatically identify key features for new applications, aiming to assist developers in requirements elicitation through machine learning and data mining techniques.

Contribution

It introduces an approach leveraging app store mining and practitioner insights to automate feature identification for new apps, supported by initial conceptual solutions and empirical interviews.

Findings

Practitioners see value in automated feature identification.

Initial conceptual solution demonstrates feasibility.

Interview study confirms need for this approach.

Abstract

App store mining has proven to be a promising technique for requirements elicitation as companies can gain valuable knowledge to maintain and evolve existing apps. However, despite first advancements in using mining techniques for requirements elicitation, little is yet known how to distill requirements for new apps based on existing (similar) solutions and how exactly practitioners would benefit from such a technique. In the proposed work, we focus on exploring information (e.g. app store data) provided by the crowd about existing solutions to identify key features of applications in a particular domain. We argue that these discovered features and other related influential aspects (e.g. ratings) can help practitioners(e.g. software developer) to identify potential key features for new applications. To support this argument, we first conducted an interview study with practitioners to understand the extent to which such an approach would find champions in practice. In this paper, we present the first results of our ongoing research in the context of a larger road-map. Our interview study confirms that practitioners see the need for our envisioned approach. Furthermore, we present an early conceptual solution to discuss the feasibility of our approach. However, this manuscript is also intended to foster discussions on the extent to which machine learning can and should be applied to elicit automated requirements on crowd generated data on different forums and to identify further collaborations in this endeavor.