The project SyKo²W² (Synthesis of Completion and Correction for Knowledge Graphs on the Web) deals with the improvement of large Web Knowledge Graphs, such as DBpedia, YAGO, or Wikidata.

    Key Idea

    In the past, a number of methods both for completing missing links in knowledge graphs, as well as for identifying wrong links have been proposed. Although both problems look like two sides of the same coin, most proposed approaches concentrate only on one of the two problems, and there are almost no methods which combine both completion and correction in knowledge graphs.

    Expected Outcomes

    While the overall aim of the project is to develop integrated methods for completion and correction in Web Knowledge Graphs, intermediate milestones include

    • A systematic benchmark dataset collection
    • A quantitative comparison of existing approaches, based on that collection
    • The exploration of pipelining approaches (i.e., serial combinations of completion and correction approaches), with and without a feedback loop
    • The development and application of model based outlier detection methods, where the model is used for predicting missing knowledge, while the outlier detection is used for finding wrong links
    • The development and application of prediction methods which identify errors as a side effect


    The project is funded by the state of Baden-Württemberg under the funding scheme for assistant professors (Juniorprofessoren-Programm), with the total funding amount totalling at 150k €.