Given a supervised/semi-supervised learning scenario where multiple annotators are available, we consider the problem of identification of adversarial or unreliable annotators.
@article{arxiv.1605.09432,
title = {Evaluating Crowdsourcing Participants in the Absence of Ground-Truth},
author = {Ramanathan Subramanian and Romer Rosales and Glenn Fung and Jennifer Dy},
journal= {arXiv preprint arXiv:1605.09432},
year = {2016}
}
Comments
4 pages, 5 figures, Workshop on Human Computation for Science and Computational Sustainability, NIPS 2012, Lake Tahoe, NV. 7 Dec 2012