A scholarly knowledge graph is only as strong as the quality of its data. Collecting metadata once and calling it complete is not enough. Research data changes constantly; authors update their profiles, funders revise project details, repositories fix errors, publishers correct article information…
Without ongoing validation, any knowledge graph quickly becomes outdated and unreliable, which is why the OpenAIRE Graph validates continuously.
Validation is not a one-time checkpoint. It is an ongoing process that keeps the OpenAIRE Graph accurate and trustworthy.
How monthly updates work
Every month, the OpenAIRE Graph reconnects with thousands of data sources and pulls in fresh metadata. But the process goes beyond simply adding new records, the entire Graph is rebuilt using enhanced AI and text mining methods and is then compared with the previous version to ensure consistency and quality.
When an author name is fixed, a missing DOI is added, or publication date is corrected by a source, those improvements appear in the next Graph update, allowing the Graph to evolve alongside the research landscape it represents. These updates are then reflected in OpenAIRE EXPLORE and other services built upon the Graph, such as OpenAIRE MONITOR and CONNECT.
Multiple layers of quality checks
To ensure metadata quality and accuracy, OpenAIRE does not rely on a single validation check-pont, but instead runs several checks in parallel. These include,
- Standards-based checks. Metadata from providers that follow OpenAIRE compatibility standards is checked against the OpenAIRE Guidelines. This ensures structural integrity, adherence to interoperability standards, and that required information is provided. Sources like Crossref, DataCite, and PubMed have different schemas, so checks are adapted accordingly.
- Managing metadata from multiple sources. When multiple repositories describe the same publication, the validation system applies trust-based rules that prioritise metadata from more authoritative sources, while preserving all variants to ensure transparency.
- Assigning trust levels to identifiers. Not all persistent identifiers carry equal weight. A DOI from Crossref is highly reliable; the same DOI manually entered by an institutional repository is less certain. Similar principles apply to ORCID and ROR identifiers.
- Full text mining for verification. Text mining of PDFs helps confirm metadata or extract missing details.
- Expert review. User-contributed links undergo manual review processes, adding a human layer to automated checks.
Example: How ORCID Verification Works
Correctly linking publications to their authors is essential for giving proper credit, making funding decisions, and performing research evaluation. But many authors share similar names, and identifiers can be entered incorrectly. A wrong ORCID can give credit to the wrong person, or prevent someone from receiving recognition for their work.
When a publication claims that Author X has ORCID Y, OpenAIRE does not immediately accept the claim. Instead, that ORCID enters a “pending” state. It is acknowledged but not yet verified.
Verification happens when OpenAIRE independently collects works directly from the ORCID registry and compares them with Graph records. If a Graph record and an ORCID work share the same DOI, and the author names match, the pending ORCID is confirmed.
OpenAIRE does not trust claimed identifiers at face value. It verifies them against authoritative sources.
A cycle that benefits everyone
Content providers can use validation reports to improve the quality of their own repositories, supported by the expertise of the OpenAIRE aggregation team. When reoccurring issues appear, they can work directly with the team to understand problems, adjust their workflows, and align with the OpenAIRE Guidelines.
This creates a positive cycle. Better, FAIR-aligned source data produces a better Graph. A better Graph produces better feedback. Better feedback leads to better source data. The whole research ecosystem benefits.
Why it matters
Research assessment, funding decisions, and scientific discovery increasingly depend on data from knowledge graphs. Institutions use them to evaluate research impact. Funders use them to track project outputs. Researchers use them to discover relevant work and potential collaborations.
For decades, these activities relied on proprietary databases that are costly and often opaque. The research community has limited insight into how data is collected, validated, or presented. Open infrastructures offer a more transparent alternative. But openness alone is not enough. An open knowledge graph full of errors, duplicates, or unverified claims can distort evaluation and even cause harm.
Continuous validation is therefore not a technical nicety. It is what turns an open collection of data into a trustworthy, reliable, and accountable infrastructure. Every monthly update, every verified ORCID, every resolved conflict, and every transparency log contributes to something bigger: an open resource that institutions, funders, and researchers can rely on for discovery, assessment, and decision-making.
Quality is what turns openness into trust.
Rigorous validation is what allows open infrastructure to become a credible alternative for bibliometrics, assessment, and discovery.
