CODECOD: Crowdsourcing Platform for Code Smell Detection

Finding code smells in a program code must be done as soon as possible to improve the software maintainability. Nowadays, various automatic code smell detection tools have been developed. However, to increase the quality, the role of humans who do manual detection is still needed. Therefore, in this work, we develop a platform called CODECEOD, which involves crowd to detect code smells. This platform implements crowdsourcing method by decomposing requested tasks, in the form of uploaded source codes, into microtasks to enable the distribution of tasks to multiple workers. To guarantee the quality of the detection results, we introduced a quality assurance method called Find, Vote, Verify. Based on the evaluation involving software engineers, CODECOD is capable to detect more code smells with high accuracy compared to an automatic tool. Moreover, we also show that the proposed Find, Vote, Verify technique delivers an improved accuracy compared to the traditional output-agreement quality assurance technique.