Doubly Robust Crowdsourcing
This work addresses the challenge of efficiently building large-scale labeled datasets for AI using crowdsourcing, offering a method to reduce costs while maintaining accuracy, though it is incremental as it builds on existing statistical estimation techniques.
The paper tackles the problem of aggregating noisy labels from crowdsourcing at test time by formulating it as a statistical estimation of expected voting scores, using a doubly robust framework with supervised learners to reduce variance. Experiments show this approach significantly lowers label costs while achieving accuracy close to an ideal scenario where all workers label all data points.
Large-scale labeled dataset is the indispensable fuel that ignites the AI revolution as we see today. Most such datasets are constructed using crowdsourcing services such as Amazon Mechanical Turk which provides noisy labels from non-experts at a fair price. The sheer size of such datasets mandates that it is only feasible to collect a few labels per data point. We formulate the problem of test-time label aggregation as a statistical estimation problem of inferring the expected voting score. By imitating workers with supervised learners and using them in a doubly robust estimation framework, we prove that the variance of estimation can be substantially reduced, even if the learner is a poor approximation. Synthetic and real-world experiments show that by combining the doubly robust approach with adaptive worker/item selection rules, we often need much lower label cost to achieve nearly the same accuracy as in the ideal world where all workers label all data points.