Understanding Android Crowdsourced Worker through Portraits

Abstract

Millions of crowdsourced workers are using crowdsourced platforms for their second jobs. They have different testing skills, styles, preferences, etc. Understanding them is important for making collaborative decisions such as crowdsourced task assignments. Existing crowdsourced platforms do not provide enough information about crowdsourced workers, and we need to spend a lot of effort searching for this information on crowdsourced platforms. In contrast to the basic worker information displayed on crowdsourced platforms, we propose describing workers as a quick way to characterize and understand them. We discuss how to build portraits of workers that are concise and informative. We propose a multidimensional model for Android crowdsourced worker portraits to specify attributes about various aspects of software testing. Then, we propose a methodology that utilizes text analytics, web data analytics, and test script analytics techniques to analyze various sources of data about workers on a crowdsourced testing platform in order to construct the portraits. The constructed portraits can be vividly displayed on the web to help people quickly learn about crowdsourced workers and make better decisions when collaborating on testing software. Results show the potential for recommended improvements and correct assignments when using our portraits. Worker portraits are an effective form of characterizing workers. It helps one to quickly understand workers and can be applied to a variety of applications in the software testing process.