Skip to main content
Version: 11.2.0

Guide to the Graph Datasets

Monthly releases vs 6-month dataset releases: what’s the difference?​

OpenAIRE offers different ways to access the Graph content, following a differentiated provisioning cadence: some access pathways are monthly, while others are released on a six-month cycle.

Monthly releases: best for “what’s happening now”​

Monthly updates give you the latest view of the Graph. These are reflected in:

  • OpenAIRE EXPLORE and other service portals like CONNECT, MONITOR, and PROVIDE.
  • Public APIs.

Six-month dataset releases: best for “a stable snapshot I can reproduce”​

The Graph dataset, and thereby the Beginner’s kit and Cloud access, are released on a six-month cadence. These allow you to:

  • Run reproducible analyses (you can cite a specific release DOI).
  • Process large volumes of records efficiently (bulk download + local processing).
  • Build offline pipelines.
  • Keep a stable snapshot for audits or reporting cycles.

Which to choose? The data access trilemma​

A useful way to understand which access pathway best fits your needs is to look at the data access trilemma of Free, Complete, and Easy.

OptionCharacteristicsDescription
The Graph datasetFree + CompleteYou get all the data, but you need storage, computational resources, and skills to process large raw files.
The Beginner’s kitFree + EasyDesigned to run locally with minimal setup and a smaller subset, so you can learn and prototype without heavy infrastructure.
Cloud accessEasy + CompleteWorks on all the data on the cloud, relying on Google BigQuery, and shifting costs to query processing rather than download/hosting.

Why not release the full dataset monthly?​

Each Graph dataset release is large and preserved over time on Zenodo. Publishing a full new dataset every month would create many repeated large-scale exports and significantly increase long-term storage requirements.

A six-month cadence is a more sustainable approach: it keeps dataset releases stable and citable, while monthly pathways remain available for users who need the latest view.

A simple rule of thumb​

  • Choose monthly pathways if you need the latest view (discovery, live monitoring, incremental integrations).
  • Choose the six-month dataset release if you need a citable, reproducible snapshot for analysis, reporting cycles, or building robust offline pipelines.

Choose the right dataset​

The datasets are available in two broad families:

  1. The Full Graph dataset (complete export, split into multiple archives).
  2. Sub-graph datasets (lighter slices tailored for specific needs).

The Full Graph dataset: what each archive contains​

Link: https://doi.org/10.5281/zenodo.20428976

The Full Graph dataset is split by entity types and relation types.

Entity archives​

These archives contain the descriptive records for different entity types:

  • publication_[part].tar — metadata about research literature
  • dataset_[part].tar — metadata about research data outputs
  • software.tar — metadata about research software
  • otherresearchproduct_[part].tar — metadata for outputs not classified as literature/data/software (e.g., physical object, image, audiovisual, etc.)
  • organization.tar — metadata about organisations (universities, research organisations, funders, etc.)
  • datasource.tar — metadata about data sources (repositories, journals, aggregators, funder databases, etc.)
  • project.tar — metadata about project grants
  • person.tar — metadata about people
  • communities_infrastructures.tar — metadata about research communities and infrastructures

Relation archives​

Relationships are provided in archives named using the pattern: [source_type][semantics][part].tar

Example: product_Cites_*.tar for citation relationships. This naming approach makes the meaning of the relation explicit (“semantics”) and supports splitting large relation sets into parts.

Sub-graph datasets: lighter, purpose-driven slices​

If you don’t need the entire full dataset, sub-graphs are designed to be useful and easier to handle for specific use cases:

  • Dataset for research communities and initiatives: A curated slice relevant for communities collaborating with OpenAIRE and initiatives with a Community Gateway on OpenAIRE CONNECT.
  • Dataset of funded products: Metadata records about research products with funding information, grouped by funder-specific archive files.
  • New collected projects dataset: Project/grant metadata collected since a reference date (delta release), useful to track changes in project coverage or newly added grants.

Where to find OpenAIRE Graph datasets​

OpenAIRE Graph datasets are available on Zenodo. You can find them by navigating Zenodo in the following steps:

  1. Start from the OpenAIRE Graph Zenodo community page and filter for datasets.
  2. Pick the release you need (each release has its own DOI, files, and statistics).
  3. Download the subset that matches your use case.

Note: Treat a Zenodo release as a versioned snapshot. If you want your work to be reproducible, always record the DOI of the release and the date/version you used.

Understand the download format​

If you download one of the archives, the structure is as follows:

  • A .tar archive (a container file)
    • Containing .gz files (compressed)
      • Containing JSON Lines (JSONL), meaning one JSON object per line.

This format is common for large-scale open datasets because it supports efficient processing (reading line-by-line rather than loading a giant array into memory).

Two essential resources:​

  1. The Schema: Defines what a “valid” record looks like.
  2. The Data Model Documentation: Explains the meaning of entities, relationships, and related fields.

Practical recommendation: Validate a small sample early. It’s the quickest way to catch mismatched assumptions before you build a full pipeline.

Getting started​

A practical workflow for accessing the Graph dataset:

  1. Define your question (what exactly do you need to analyse?).
  2. Choose the smallest dataset that answers it (prefer sub-graphs when possible).
  3. Download from Zenodo and record the release DOI/version.
  4. Unpack: tar →\to gz →\to JSONL.
  5. Inspect and validate a small sample against the schema.
  6. Process and load into your tool of choice (SQL/Python, database, graph DB, etc.).
  7. Document and cite the dataset DOI in outputs to support reproducibility.

Reuse and attribution​

OpenAIRE Graph datasets are released under CC BY 4.0, supporting reuse with attribution. When publishing results, it’s good practice to cite:

  • The Zenodo release DOI you used.
  • The release version/date, plus a short description of your processing steps.

This makes your work easier to reproduce and trust.