Human-Powered Blocking in Entity Resolution: A Feasibility Study

Entity Resolution (ER) is the problem of matching the records that refer to the same entity within or across two or more data sources. In recent years, human-powered ER solutions have been proposed so that challenging ER tasks, that machines cannot do well, can be helped by human workers. While successful in achieving high matching accuracy, existing human-powered ER methods did not incorporate a core technique, i.e., blocking, for improving the scalability of the ER process. To address this issue, this paper carries out the feasibility study to validate whether the blocking technique can be integrated into the human-powered ER. Specifically, we first propose two variations of human-powered blocking methods. We then validate their effectiveness in improving the scalability of the ER process through simulated crowdsourcing and AMT-based experiments in synthetic and real-life datasets, respectively.