Keynote Talk
"The Structure Behind the Answer"
Abstract: Knowledge graphs support applications ranging from data integration to AI, but their flexibility raises challenges for data quality and the interpretation of query results. Schemas may be implicit, repeated facts can diverge during updates, and unexpected answers can be difficult to explain. This talk examines how constraints, normalization, and query provenance help address these challenges. Validating shapes can be extracted from large graphs, using support and confidence to assess candidate constraints and filter out spurious patterns. Functional dependencies within and across nodes and edges guide normalization to reduce redundancies that can lead to inconsistencies during updates. Why-provenance identifies the nodes, edges, labels, and properties contributing to a particular query answer. This provenance can be computed through query rewriting and evaluation on existing graph database engines, making it possible to inspect the specific data behind an answer.
Biography
Katja Hose is Professor of Data Management at TU Wien, where she leads the Data Management and Knowledge-Driven AI (DMKI) Lab. Her research focuses on graph data management, knowledge graphs, query optimization, data quality, and knowledge-driven AI. She serves as program co-chair of VLDB 2027 and previously co-chaired EDBT 2023, ESWC 2021, and the research track of ISWC 2024. She is on the editorial boards of the VLDB Journal, Transactions on Graph Data and Knowledge, and Semantic Web. Before joining TU Wien, she was a professor at Aalborg University and a postdoctoral researcher at the Max Planck Institute for Informatics. She holds a PhD from Ilmenau University of Technology.