High Agreement Block

AbstractThis article outlines a methodology that uses crowdsourcing to reduce the workload of experts for complex semantic tasks. We split turker-annotated datasets into a high-agreement block, which is not modified, …...