Rock: Cleaning Data by Embedding ML in Logic Rules

• DOI: 10.1145/3626246.3653372
• Publication Date: 2024-05-23T14:26:39Z
• Authors: Xianchun Bao; Zian Bao; Bie Binbin; QingSong Duan; Wenfei Fan; Hui Lei; Daji Li; Wei Lin; Peng Liu; Zhicong Lv; Mingliang Ouyang; Shuai Tang; Yaoshu Wang; Qiyuan Wei; Min Xie; Jing Zhang; Xin Zhang; Runxiao Zhao; Shuping Zhou

We introduce Rock, a system for cleaning relational data. Rock implements framework that unifies machine learning (ML) and logic deduction by embedding ML classifiers in rules as predicates. In unified process, it identifies tuples refer to the same real-world entity, catches semantic inconsistencies among entities, deduces timeliness of attribute values imputes missing possibly extracting data from knowledge graphs. That is, conducts entity resolution, conflict incomplete information imputation makes use their interactions improves overall quality Moreover, supports methods, batch incremental, discovering real-life data, detecting errors with learned rules, accumulating ground truth, fixing errors, such corrections are logical consequences truth. present design implementation Rock. evaluate scalability accuracy share lessons variety applications.