How do we integrate community data from Wikidata and the fuzzy-sl Wikibase into a cultural heritage knowledge graph?

Community-curated Wikibase ecosystems, most notably Wikidata, FactGrid, and wikibase.cloud instances such as fuzzy-sl, have become significant sources of Cultural Heritage (CH) and archaeological data. In parallel, infrastructure initiatives (e.g., NFDI4Objects) are building CIDOC-aligned knowledge graphs that demand robust provenance, semantic interoperability, and reproducibility. This paper presents a semi-automated workflow for integrating Wikibase data into infrastructure-scale graphs, including entity selection, ontology design aligned with CIDOC CRM (with CRMarchaeo, CRMsci, and CRMdig as needed), scripted RDF transformation, open publication of code and data, versioned snapshot releases, and ingestion into the NFDI4Objects Knowledge Graph. The approach preserves Wikidata-style qualifiers and references while yielding an event- and provenance-centric representation. Two use cases demonstrate feasibility and limits. The Irish Holy Wells dataset showcases richly reified statements (e.g., use, sources) mapped to CIDOC events. The Campanian Ignimbrite findspots in fuzzy-sl focus on location-centric modelling and interdisciplinarity, requiring E53 Place as baseline with CRMarchaeo/CRMsci specialisation for stratigraphy, sampling, and analysis. We discuss challenges in ontology alignment, granularity, explicit treatment of uncertainty (“fuzzy/wobbly” data), and sustaining semi-automated pipelines amid evolving community schemas. We argue for identifier discipline, machine-actionable provenance, and FAIR Digital Objects for each release. The outcome is an interoperable, federated ecosystem, spanning triplestores, Wikibases, and FDOs, in which community knowledge bases become partners of infrastructure graphs.