ReDQ Rethinking Data Quality for generative AI and knowledge graphs

A Dagstuhl-like workshop at The 25th International Semantic Web Conference (ISWC) 2026.

Description

Knowledge Graphs (KGs) are increasingly relied upon as semantic infrastructure for AI systems, they act as the foundation for generative models, support retrieval, and serve as machine-actionable memory for autonomous agents. This diverse use places new demands on KGs that traditional data quality (DQ) dimensions, such as accuracy, completeness, consistency, timeliness, only partially capture: what matters is no longer just whether facts are correct, but whether a KG’s semantics are sound enough for an AI system to depend on, including well-formed ontological commitments, correctly disambiguated entities and relations, traceable provenance, and machine-interpretable meaning. A KG can score well on conventional DQ metrics and still fail as AI infrastructure, since classical frameworks were designed for static, human-facing data rather than knowledge that machines interpret, traverse, and reason over. This Dagstuhl-style workshop argues that AI usage surfaces a new DQ landscape specific to KGs as semantic artifacts, distinct from DQ for AI in general, and brings together researchers and practitioners from the Semantic Web, Knowledge Graph, data management, and AI communities to identify the DQ dimensions distinctive to KGs in AI contexts and develop a shared research agenda going forward.

Topics

The workshop focuses on the semantic dimensions of DQ that emerge when KGs serve as infrastructure for AI systems, distinguishing it from DQ assessment in general or AI applications more broadly.

Topics of interest include:

• Redefining DQ in the AI era: Which traditional quality dimensions remain relevant, and which new dimensions are needed?
• Semantic DQ dimensions for KGs in AI contexts: ontological soundness, entity/relation disambiguation, and schema alignment, alongside provenance, explainability, representativeness, uncertainty, and contextual relevance.
• Measuring and operationalizing AI-oriented DQ: Metrics, assessment methodologies, and quality metadata.
• Towards a shared framework: Taxonomies, benchmarks, and community recommendations for AI-ready DQ.

Date and location

Registration

If you would be interested in taking part, we kindly ask you to complete the following Google Form to express your interest. The responses will help us identify potential participants for the workshop.

Further information is also available on a Padlet.

Organizing committee

• 

  Lorena Etcheverry

• 

  Lisa Ehrlinger

• 

  Katja Hose

• 

  Maria Angela Pellegrino

• 

  Anisa Rula

Program

Contact information