Threefold similarity analysis: a case study on crowdsourcing feeds