Knowledge Graph Repair

Knowledge Graph Repair

연구 분야: Databases

학회: European Semantic Web Conference

The Semantic Web provides standards for knowledge graphs (KGs), which have become popular for solving data heterogeneity problems in enterprises, since they allow for flexible data modelling and integration via linking of graphs. This flexibility requires standards to ensure data quality and consistent state, such as the Shapes Constraint Language SHACL. However, SHACL does not provide the means to explain why constraint violations occur and how the KG can be repaired to conform to the constraints. Also, repairs for a KG can come with a high number of different alternative choices to pick from, where we need a way to determine preferences in practice. Finally, knowledge in the KG itself can be exploited for repairs to determine fresh values and preferred choices and to identify incorrect data from a real-world perspective. For these challenges, we aim to develop a system that combines logic-based repairs and data-driven analysis for a repair approach that concludes KGs towards a quality fix point. The approach will not only be defined at the formal level, but we also will provide prototypical implementations for practical experiments, thereby positioning it at the intersection of theoretical and applied research. Use cases shall be provided from companies, projects and open data to better understand how repairs can be applied effectively in practice. With this work we contribute to improving the quality of KGs by providing intelligent knowledge graph repair.