Motivation
KGs are central infrastructures for semantic interoperability, data integration, and structured knowledge representation across scientific, cultural, and industrial domains. However, many KG formalisms and engineering practices still treat statements as if their validity were independent of the conditions under which they are produced, interpreted, accepted, revised, or queried. This assumption becomes limiting in domains where knowledge is contextual, evolving, contested, or perspectival, and where statements may depend on provenance, temporal scope, epistemic status, methodological assumptions, community viewpoints, or changing conceptual frameworks.
The Semantic Web already offers several mechanisms for contextualising statements — RDF Reification, N-ary Relations, Named Graphs, RDF-star / RDF 1.2, provenance vocabularies such as PROV-O, Nanopublications, Wikidata-style qualifiers, Singleton Properties, Temporal Knowledge Graphs, fluents-based models, and specific semantics such as Conjectures. These provide important building blocks but differ in granularity, semantics, and operational assumptions. The open challenge is not only how to attach metadata to statements, but how to represent, query, and reason over the conditions under which statements are valid, applicable, contested, superseded, perspectival, or meaningful across multiple heterogeneous dimensions. A Zotero group will be created for collaborative bibliography management.
Timeliness
Although contextual and multi-perspective modelling has long been considered in Knowledge Engineering, existing solutions often remain local, pattern-based, or tied to specific modelling tasks. A community-wide re-evaluation is timely because KGs increasingly ground neuro-symbolic systems and Retrieval-Augmented Generation, while current symbolic representations still struggle to expose provenance boundaries, temporal scope, disagreement, and concept drift. The community addresses related problems through fragmented notions such as contexts, perspectives, layers, lenses, views, and possible worlds. MKGs offer an umbrella concept for comparing these strands and identifying the trade-off between formal expressivity and scalable storage/query execution.
Why a Barstuhl format
This agenda is best developed through an interactive format rather than standard paper presentations. The workshop uses lightning talks, breakouts, and plenary synthesis to bring foundational, database, neuro-symbolic, and domain perspectives into direct dialogue and to produce a shared roadmap for future MKG research.
Objectives & expected outcomes
The workshop serves as a collaborative incubator to produce a community-drafted manifesto and framework document. This document is the immediate foundation for a co-authored White Paper on Multi-dimensional Knowledge Graphs, capturing a unified set of research questions, concrete engineering challenges, and pragmatic, cross-domain reference use cases. It will be shared through the workshop website, Zenodo, and/or an open collaborative repository, and may form the basis for a future Dagstuhl seminar proposal.
Manifesto / framework document
A collaboratively drafted framing document capturing the workshop's discussions and points of consensus.
MKG White Paper
A more developed white paper building on the manifesto, intended for wider community dissemination.
Possible Dagstuhl seminar
A potential follow-up Dagstuhl seminar proposal, continuing the discussion in a dedicated venue.
Outcomes will be disseminated via this website, Zenodo, and/or an open repository.
Research questions & engineering challenges
Two interlocking tracks — formal foundations and computational engineering — frame the workshop's breakout sessions.
What minimal, sound formalisms can natively represent multi-dimensional assertions (time, perspective, stance) while preserving W3C backward compatibility?
How can we align divergent, contradictory dimensions or conflicting ontologies within the same graph without triggering logical inconsistency?
What formalisms let a model track and query the historical evolution of a concept's meaning over time without schema fragmentation?
How should query languages, storage models, and indexing strategies evolve to efficiently support querying over multiple contextual dimensions?
How can storage engines efficiently manage temporal evolution, concept drift, and multiple graph versions without excessive data duplication or query overhead?
How can multi-dimensional KGs provide a scalable symbolic foundation for neuro-symbolic AI, Retrieval-Augmented Generation, and other AI systems while preserving contextual semantics?
Breakout working groups
- (a) Formal models
- (b) Query, storage & indexing
- (c) Benchmarks & use cases