Ayato Watanabe presented his research findings at IEEE HMData 2021, a workshop affiliated with the IEEE International Conference on Big Data 2021.
The title of his paper is “Spammer Detection Based on Task Completion Time Variation in Crowdsourcing,” which focuses on a method for detecting spammers in crowdsourcing services. Crowdsourcing, which allows an unspecified large number of workers to complete tasks via the internet, faces a common issue—some workers submit low-quality outputs without seriously engaging in the tasks, often referred to as “spammers.” Various methods have already been proposed to distinguish between diligent workers and spammers. However, Watanabe’s research specifically examines the time each worker spends on labeling images in a multi-class classification task to detect spammers.
A simple detection method that labels workers with short response times as spammers can be easily bypassed if spammers intentionally delay their responses. To address this, Watanabe’s research leverages the idea that diligent workers tend to spend less time on easy classifications and more time on difficult ones, whereas spammers’ response times show little correlation with the difficulty of the classification task. His study analyzes workers’ response times based on this observation.
Originally, the workshop was planned to be held in Florida, but due to COVID-19, it was conducted entirely online. The format involved pre-recorded presentation videos, followed by a live Q&A session. Notably, the event took advantage of online-specific benefits, such as real-time voice translation and text-based question submissions via chat. However, compared to an in-person conference, opportunities for in-depth discussions and networking among researchers were somewhat limited.