As the volumes of AI problems involving human knowledge are likely to soar, crowdsourcing has become essential in a wide range of world-wide-web applications. One of the biggest challenges of crowdsourcing is aggregating the answers collected from the crowd since the workers might have wide-ranging levels of expertise. In order to tackle this challenge, many aggregation techniques have been proposed. These techniques, however, have never been compared and analyzed under the same setting, rendering a 'right' choice for a particular application very difficult. Addressing this problem, this paper presents a benchmark that offers a comprehensive empirical study on the performance comparison of the aggregation techniques. Specifically, we integrated several state-of-the-art methods in a comparable manner, and measured various performance metrics with our benchmark, including computation time, accuracy, robustness to spammers, and adaptivity to multi-labeling. We then provide in-depth analysis of benchmarking results, obtained by simulating the crowdsourcing process with different types of workers. We believe that the findings from the benchmark will be able to serve as a practical guideline for crowdsourcing applications.
展开▼
