Result description
The LUMEN Data Mesh Architecture Framework and Data Model define how four scientific communities (SSH, Mathematics, Earth Systems, Molecular Dynamics) can share and discover each other’s resources without giving up control over their own data, platforms, or internal practices. Each community packages its scientific resources (publications, datasets, software, semantic artefacts) as Data Products exposed to the federation through standardised protocols (OAI-PMH, REST, SPARQL…). The approach is decentralised and flexible: communities choose how they structure their data internally, and only need to describe what they expose using a shared, minimal metadata profile based on DCAT and DPROD. Each Data Product is accompanied by a Data Contract, a machine-readable agreement based on ODCS that specifies access conditions, update frequency, licensing, and responsibilities. No central authority imposes schemas or formats. Communities align on what matters for cross-domain discovery, and keep everything else under their own governance. This is what enables the LUMEN services to operate across domains. The Meta-Search engine reads the Data Contracts to harvest metadata from community platforms (GoTriple, zbMATH Open, Centre Mersenne, EarthPortal, MDverse), builds a shared semantic index, and supports cross-domain queries. The White-Label Discovery Platform uses the Data Model to expose resources consistently across community portals while adapting to domain-specific needs. The Software Catalogue aggregates research software metadata from multiple communities using the same Data Product and Data Contract mechanisms. To guarantee a shared semantic layer, LUMIS provides a collaborative environment for the development of FAIR-by-design semantic artefacts supporting their whole lifecycle (from creation to publication and maintenance). The governance layer is organised as a guild: representatives from each community agree on shared policies for data sharing, onboarding, and compliance, while each domain retains full autonomy over its internal practices. The framework addresses all four layers of the EOSC Interoperability Framework: technical interoperability through standard APIs and protocols, semantic interoperability through the use of DCAT and DPROD as a shared description layer for Data Products and the development of LUMIS, a platform to support the whole life cycle of semantic artefact including publication and mappings between each others, organisational interoperability through the guild-based governance, and legal interoperability through access and licensing conditions encoded in Data Contracts.
Problem addressed
Scientific resources are scattered across disciplinary platforms that were built independently, each with its own metadata schemas, access policies, and search interfaces. A researcher working on a cross-domain question has no practical way to query across communities. If someone needs to combine molecular simulation data, ocean circulation models, and social science publications on health policy, they have to visit each catalogue separately, learn each interface, and manually try to connect results that no shared infrastructure links together. Previous approaches to this problem have mostly relied on centralisation: building a single platform that aggregates metadata from multiple sources into one catalogue. This works up to a point, but it creates well-known issues. Communities lose control over how their resources are described and accessed. The central system has to accommodate every domain’s specificities, which makes it complex and fragile. And governance is unclear: who decides what metadata standard to use, who maintains quality, who is accountable when a dataset is outdated or access conditions change. The EOSC Interoperability Framework provides solid principles organised around four layers (technical, semantic, organisational, legal), and several projects are contributing to their implementation. What remains challenging in practice is giving communities a lightweight, adoptable pattern that they can use to participate in a federation without rebuilding their infrastructure or giving up their autonomy. The LUMEN Data Mesh is one concrete contribution to this effort. Instead of centralising data, it keeps ownership with the communities and connects them through Data Products and Data Contracts. Each community describes what it exposes using a minimal shared profile (DCAT/DPROD), and formalises the conditions under which it is shared (ODCS). Services like the Meta-Search, the White-Label platforms, LUMIS and the Software Catalogue then operate on top of this layer. Communities join progressively, at their own pace, without being forced to change their internal tools or practices.
