Hike: A Hybrid Human-Machine Method for Entity Alignment in Large-Scale Knowledge Bases

摘要: With the vigorous development of World Wide Web, many large-scale knowledge bases (KBs) have been generated. To improve coverage KBs, an important task is to integrate heterogeneous KBs. Several automatic alignment methods proposed which achieve considerable success. However, due inconsistency and uncertainty techniques for KBs low quality (especially recall). Thanks open crowdsourcing platforms, we can harness crowd quality. this goal, in paper propose a novel hybrid human-machine framework KB integration. We rst partition entities different into smaller blocks based on their relations. then construct partial order these partitions develop inference model crowdsources set tasks infers answers other crowdsourced tasks. Next formulate question selection problem, which, given monetary budget B, selects B maximize number inferred prove that problem NP-hard greedy algorithms address with approximation ratio 1--1/e. Our experiments real-world datasets indicate our method improves outperforms state-of-the-art approaches.