CrowdGame: A Game-Based Crowdsourcing System for Cost-Effective Data Labeling

摘要: Large-scale data labeling has become a major bottleneck for many applications, such as machine learning and integration. This paper presents CrowdGame, crowdsourcing system that harnesses the crowd to gather labels in cost-effective way. CrowdGame focuses on generating high-quality rules largely reduce cost while preserving quality. It first generates candidate rules, then devises game-based approach select with high coverage accuracy. applies generated effective labeling. We have implemented provided user-friendly interface users deploy their applications. will demonstrate two representative scenarios, entity matching relation extraction.