Component-based crowdsourcing in software engineering -

Abstract

Crowdsourcing in software engineering is a fast-growing and promising area, yet it still lacks a rigorous platform that circumvents the well-known issues such as task decomposition and composition, scheduling, coordination, quality, payment and time-to-delivery. This thesis introduces novel methods and tools for automating the crowdsourcing process of a software project, following the component-based paradigm. We mainly exploit the abilities of a professional crowd on a user-friendly and easy-to-use environment, to design, verify, code, test and then assemble various high-quality software components. The procedure starts with a request consisting of a project idea or a high-level specification on our platform. Then, the request is hierarchically developed with the help of the crowd. At each phase, the crowd competes to decompose a specification into (1) components with well-defined interfaces and (2) a glue that defines the corresponding composition operator between components. This phase is repeated until we reach atomic components that are ready for development. Additionally, the crowd competes to develop atomic components and to compose-integrate components. Our platform provides a fair and rigorous payment that can be adapted according to the requester needs as well as an intelligent rating system that mimics the crowds' performance. We evaluate our approach on non-trivial case studies and compare it to our main competitors: (1) software companies; (2) freelancers; (3) state-of-the-art software crowdsourcing platforms. Experimental results show the effectiveness of our platform with respect to cost, time-to-deliver, fairness, reusability, and quality.