Crowdsourcing semantic content: A model and two applications