MELODI: Modular Ecosystem for Linked Open Data Infrastructure

MELODI adds Linked Open Data services to the research data repositories that researchers already use. A researcher turns a dataset into a knowledge graph in one guided workflow, the repository publishes it with a DOI, FAIR metadata and long-term preservation, and anyone can query it.

MELODI is a project proposal led by Delft University of Technology for the Open Science NL Infrastructure call 2026. The letter of intent was submitted in November 2026 and the full proposal follows in April 2027.

Watch the presentation How it works

Why MELODI

Linked Open Data (LOD) is one of the most effective ways of putting open science into practice. It turns isolated datasets into a web of machine-readable, interoperable knowledge that can be queried and connected across repositories, domains and disciplines, making every dataset FAIR. Yet researchers who want to publish LOD face a persistent gap.

  • No national infrastructure

    The Netherlands has no public, national infrastructure for publishing Linked Open Data. Each group that wants to publish a knowledge graph starts from scratch.

  • Repositories stop short

    Research data repositories are not designed to support LOD. They preserve the files, but cannot deliver the deep semantic interoperability that linked data enables.

  • Build it or buy it

    Institutions either develop their own infrastructure or rely on proprietary platforms. Both routes carry serious sustainability risks once a project ends.

What MELODI does

MELODI is an open-source platform that adds LOD services to existing research data repositories without major changes to their infrastructure and data practices. Built on open standards, it runs next to the repository and connects to it through an API and a plugin, the MELODI Bridge, that links the platform and the repository software.

  • One guided workflow

    MELODI Composer, a user-centred web interface, guides researchers step by step through selecting or authoring ontologies, mapping and transforming data, and publishing a knowledge graph, with established open-source tools integrated into a single workflow.

  • Published like any dataset

    Each knowledge graph receives a DOI, FAIR metadata and long-term preservation following the repository's usual workflow, and appears beside the dataset it was made from.

  • Open to query

    Anyone can query the data through a SPARQL endpoint, and repository users see linked data views from the repository they already know.

  • A catalogue of graphs

    Every MELODI instance hosts a DCAT-AP catalogue of its knowledge graphs, compatible with EOSC and other European infrastructure.

Architecture diagram. On the left, a research data repository with its users, its servers and storage. Next to it, a MELODI instance in a portable container layer: MELODI Composer with tools for ontology editing and review, mapping and conversion, validation, persistent identifiers and reconciliation; the MELODI Bridge connecting to the repository; an API; and a triple store holding the knowledge graphs and a DCAT catalogue. Researchers publish through Composer, AI agents and reusers query through SPARQL, and other MELODI instances connect through the API.
The MELODI architecture: the platform runs next to a research data repository and connects to it through the MELODI Bridge.

From dataset to knowledge graph

Publishing linked data today needs specialised skills and a chain of loosely connected tools. MELODI Composer puts the chain into one path, with a role for everyone involved.

  1. Researchers

    Open MELODI Composer from the dataset upload, choose or create ontologies, map and convert the data, check the quality report and submit the knowledge graph for publication.

  2. Ontology engineers

    Advise on ontologies and review them inside the same workflow, instead of by email and spreadsheet.

  3. Repository curators

    Check metadata, licence and personal data, then publish the graph with a DOI, as for any other dataset.

  4. Data reusers

    Find the graph in the catalogue, then explore, query and cite it from the repository, a SPARQL endpoint or the MELODI Orchestra.

Workflow diagram in four lanes. Researchers open the dataset upload and MELODI Composer, choose or create ontologies, map and convert the data, check the quality report and submit the graph for publication. Ontology engineers advise on ontologies. Repository curators check metadata, licence and personal data and publish with a DOI. Data reusers find the graph, then explore, query and cite it.
The MELODI workflow, from a dataset upload to a published and reusable knowledge graph.

Pilots in three domains

MELODI is domain-agnostic. Three data pilots put it to work with real repositories and real datasets, publishing knowledge graphs from new and previously deposited data.

  • Natural and engineering sciences

    4TU.ResearchData, the national research data repository for the natural and engineering sciences, hosts the first MELODI instance and uses it to provide LOD services, with the TDCC-NES ontology engineers making it a cornerstone of their research support.

  • Social sciences and humanities

    A pilot with a national repository for the social sciences and humanities, to be announced with the full proposal.

  • Life sciences and health

    A pilot in the life sciences and health domain, with a partner to be announced with the full proposal.

The MELODI Orchestra

Every participating repository joins a federated network, the MELODI Orchestra, in which all knowledge graphs can be discovered and queried together. A question about materials, manuscripts and measurements no longer stops at the border of one repository.

The resulting ecosystem improves interoperability for human users and AI agents alike, and opens new avenues for interdisciplinary research.

Who is behind MELODI

  • Delft University of Technology

    Lead applicant, taking part through two units based at the university: 4TU.ResearchData, which hosts the first MELODI instance, and the TDCC-NES ontology engineers, who gathered the requirements for MELODI.

  • 4TU.ResearchData

    The national research data repository for the natural and engineering sciences hosts the first MELODI instance and offers LOD services with it.

  • TDCC-NES

    The Thematic Digital Competence Centre for the natural and engineering sciences. Its ontology engineering service supports researchers across the Netherlands and will make MELODI a cornerstone of that support.

Further partners in the Netherlands and abroad are completing the consortium before the full proposal.

Duration
36 months
Budget
€1.7 million
Work packages
5
Data pilots
3

The work is organised in five work packages: coordination, governance and sustainability; core infrastructure and repository integration; MELODI Composer and tools; catalogue and federation; pilots, training and community building.

Watch the presentation

A 15-minute presentation of the MELODI proposal, recorded in October 2026.

Timeline

  1. Letter of intent submitted to Open Science NL. ✓
  2. Preparatory workshop for main applicants in Utrecht.
  3. Deadline for the full proposal.
  4. Funding decision by the Open Science NL steering board.
  5. Start of the 36-month project, if funded.

The call is described on the Open Science NL website.

Frequently asked questions

Who developed the MELODI proposal?

The TDCC-NES ontology engineers and the 4TU.ResearchData team.

Where will the proposal be submitted?

To the Open Science NL Infrastructure call. We will request €1.7 million for a project duration of three years. The application deadline is 13 April 2027 and a final decision is expected by December 2027.

Who are the other partners?

4TU.ResearchData is the primary partner and will be the main applicant. We have expressions of interest from multiple Dutch institutions (co-applicants) and international partners (listed as third parties), and we also expect to reach agreement with a second research data repository.

Why partner with repositories?

Because they have a mission to curate and preserve data in the long term; they have a direct relationship with researchers, data stewards and data reusers; and they hold datasets that could be readily transformed into LOD.

Can MELODI function without being attached to an existing repository?

Yes, it is fully open-source and does not require the repository to function. Anyone will be able to freely deploy a standalone instance of MELODI.

Will MELODI be cross-domain?

Yes, and to demonstrate the potential of cross-domain LOD infrastructure, we plan to develop three data pilots covering NES, SSH and LSH, with the help of partners from each domain.

Are the other TDCCs involved?

We have started discussions with TDCC-SSH and plan to do the same with TDCC-LSH. We would love to have them both on board.

What about domain-specific open data infrastructures?

They are welcome to join the data pilots and we expect them to play a bigger role at a later stage, after the initial prototype of MELODI is deployed.

How does MELODI connect to EOSC and other EU-wide infrastructure?

We look forward to that, but it is too early. First we need to demonstrate the viability of the prototype.

Will any researcher be able to use MELODI?

Yes, MELODI is user-centred and its primary goal is to address the needs of researchers and data stewards.

Which services will MELODI provide to researchers?

By the end of the project, it will offer a complete workflow from ontology selection to data mapping and transformation to LOD storage and querying. Importantly, MELODI will not develop tools, but rely on existing open-source tools and on software developed by TDCC-NES through separate funding sources. The tools will be integrated into a web interface called MELODI Composer.

What about performance?

MELODI is designed to provide efficient and scalable LOD storage, querying and federation via the QLever triple store.

Will MELODI be provided as a service to research institutions?

The MELODI prototype is hosted and managed by research data repositories. In the initial phase we do not plan to provide MELODI as a service, but institutions will be able to deploy their LOD via repositories, and also potentially deploy their own local instances.

What about legacy LOD infrastructure?

We will encourage institutions with legacy infrastructure to migrate to MELODI and will make it as easy as possible to do so. MELODI will be interoperable with legacy infrastructure via standard SPARQL APIs.