Continuous Knowledge Graph Quality Assessment through Comparison using ABECTO

Abstract

Maintaining accuracy and completeness of RDF knowledge graphs is an important but challenging task. As constraints checking can only spot outliers, knowledge graphs would need to be checked against reference data to obtain a reliable assessment. But this is often impossible due to the lack of suitable reference data. An alternative is the comparison with other, overlapping knowledge graphs that possibly contain incorrect data, too. This can spot potentially incorrect values in the maintained knowledge graph. As knowledge graphs might evolve over time, this comparison must be done regularly. However, the regular comparison of data is an exhausting and error prone task, if done manually. We present ABECTO, a command line tool for the automatic comparison of multiple possibly incorrect RDF knowledge graphs to monitor their accuracy and completeness. In our demonstration, we will showcase its application in a Continuous Integration scenario for the regular automatic check of the quality of a maintained knowledge graph.