Request for comment: Metadata enrichments in practice

[codycooperross]

DataCite is developing community-driven DOI metadata enrichment processes as part of an ongoing, multi-faceted effort to improve the quality and utility of the DataCite metadata store and enhance discovery, reuse, and reporting across the research ecosystem. You can learn more about our previous explorations in this space, including our high-level goals and priorities, in our earlier request for comment. Below, we share our latest work on enrichments and invite ongoing community feedback as we continue to iterate on our approach.

Sharing an initial set of enrichments

We are excited to release our first set of enrichment records produced through a collaboration with COMET (Collaborative Metadata) and available at a new enrichments endpoint of the DataCite REST API: https://api.datacite.org/enrichments. At this time, we have ingested 26.9 million enrichment records derived from two COMET enrichment projects. The first project maps affiliation strings to Research Organization Registry (ROR) identifiers, establishing 26.1 million new connections between creators/contributors and ROR ID affiliations. The second links arXiv preprints to their published versions, creating 855,764 new connections between previously unlinked resources. In combination, these two projects enrich over 6.3 million unique DOI records.

Each individual enrichment record includes provenance metadata connecting the record to both its contributors and the resources that the enrichment was derived from, such as project documentation and related datasets. You can learn more about our approach to provenance and the overall structure of enrichment records in our previous request for comment.

REST API tools are also available for exploring enrichments by DataCite repository and by DOI. Retrieve enrichments for a specific repository’s DOIs by entering a Repository ID in the client-id= parameter of the enrichments endpoint, e.g., https://api.datacite.org/enrichments?client-id=arxiv.content. Explore enrichments for a specific DOI by entering the DOI name in the doi= parameter of the enrichments endpoint, e.g., https://api.datacite.org/enrichments?doi=10.60692/44wdx-8kd18.

Previewing enriched DataCite DOI records

Enrichment records act as a non-destructive layer over original DOI records. When combined with original DOI records, they can be used to generate separate, enriched DOI records. We’re excited to share a preview of enriched DOI metadata records, enhanced with the ingested enrichments described above.

Enriched DOI metadata records can be accessed with a new enriched=true flag in the REST API for single DOI records with associated enrichments. For example, try: https://api.datacite.org/dois/10.60692/44wdx-8kd18?affiliation=true&publisher=true&enriched=true. This DOI (10.60692/44wdx-8kd18) has 60 associated enrichment records, each listed in the relationships section of the API response. When enriched=true, the associated enrichment records are applied non-destructively to the original record, producing an enriched DOI record with 67 affiliation connections to ROR IDs where there were none originally. The original DOI metadata record is stored separately from the enriched record and can always be accessed by omitting the enriched=true parameter; e.g., https://api.datacite.org/dois/10.60692/44wdx-8kd18?affiliation=true&publisher=true.

The process of applying enrichment records is responsive to changes in the original record and prioritizes metadata validity. For example, enrichment records are not applied if targeted metadata fields are changed by the original depositor, and the original DOI record is displayed in place of the enriched record if the enriched record is in any way invalid according to the DataCite Metadata Schema.

Tip

Learn more about working with these initial metadata enrichment features in our Support documentation: https://support.datacite.org/docs/metadata-enrichments

Next steps

We are collecting community feedback as we expand our work on enrichments and explore enrichment curation tools and workflows for the metadata creator community. We invite you to share input via our new enrichments survey, by responding to this post, or by sending us an email at support@datacite.org. Input is welcome any time, and we will continue to keep this discussion updated as we continue development. Thank you for participating!

[tkkuehn]

(For credit's sake, I wrote this in collaboration with my colleague John Aspler who is not on GitHub)

This is very exciting -- I see a lot of potential in this work (in particular having so many new ROR affiliations in place could make a big difference to Canadian data discovery in Lunaris). That said, to use these enrichments, I need to be able to systematically access the ones that are relevant to my use case, and ideally I would have some way to assess their correctness.

As far as I can tell, using the DataCite API to query across the content of these enrichments is somewhere between fiddly and impossible right now (short of working with the datasets COMET has published on Hugging Face). On the one hand, it would be nice to have the enriched=true filter function for multi-DOI queries, but on the other hand, without having the stewards of those DOIs' metadata vet their correctness, it's hard to know what risks are involved (especially because which of the two COMET enrichment projects an individual enrichment is associated with does not appear to be encoded in that enrichments' metadata; maybe a RAiD use case?). Making the great contribution embodied by these enrichments available without overcomplicating the APIs seems like a challenge.

Ideally, there would be a pathway for enrichments to be applied to original/canonical DOIs when they are correct, but I am not sure if there is a practical model for making those assessments or involving DOI stewards in that decision making. A clear vision for how metadata stewards, metadata consumers, and related infrastructure providers (e.g. ORCID) are meant to use these enrichments in practice would be very helpful!

Finally, I'm curious if these enrichments are intended to always come from relatively trusted projects (presumably DataCite was involved or at least reviewed the COMET projects before deploying the enrichments). Will community projects be possible? Will there be mechanisms in place to control who can submit what kinds of enrichments - or to indicate the source of an enrichment? What role might LLM-powered systems play?

[codycooperross]

Thank you so much for this thoughtful response, @tkkuehn ! We really appreciate the detailed feedback.

That said, to use these enrichments, I need to be able to systematically access the ones that are relevant to my use case, and ideally I would have some way to assess their correctness.[...]As far as I can tell, using the DataCite API to query across the content of these enrichments is somewhere between fiddly and impossible right now (short of working with the datasets COMET has published on Hugging Face).

The enrichments endpoint (https://api.datacite.org/enrichments) currently supports two filters: DataCite repository ID (client-id=) and DOI (doi=). Can you describe some of the other ways you might want to find enrichments relevant to you and what you might want to do with those enrichments afterward?

On the one hand, it would be nice to have the enriched=true filter function for multi-DOI queries, but on the other hand, without having the stewards of those DOIs’ metadata vet their correctness, it’s hard to know what risks are involved (especially because which of the two COMET enrichment projects an individual enrichment is associated with does not appear to be encoded in that enrichments’ metadata; maybe a RAiD use case?). Making the great contribution embodied by these enrichments available without overcomplicating the APIs seems like a challenge.

We’re exploring work on the enriched=true filter for multi-DOI queries alongside other developments involving metadata steward curation, mentioned below. While each enrichment record includes provenance metadata containing connections to resources that describe how the enrichment was produced (at the moment, Hugging Face pages with datasets and project documentation as well as project DOIs in some cases), I can understand via your suggestion of RAiDs that a singular, PID-identified resource might be an easier way to examine and assess enrichments by project. Let me know if that doesn’t ring true or if you’re seeing other challenges as well. Here is an example of an enrichment with a project DOI as a resource: https://api.datacite.org/enrichments/f670c88c-3f9a-4740-84e8-fa6ed46c200d

Ideally, there would be a pathway for enrichments to be applied to original/canonical DOIs when they are correct, but I am not sure if there is a practical model for making those assessments or involving DOI stewards in that decision making. A clear vision for how metadata stewards, metadata consumers, and related infrastructure providers (e.g. ORCID) are meant to use these enrichments in practice would be very helpful!

This is a valuable assessment! Our next phase of work is to support the curation and decision-making processes you describe as we continue to iterate and build a metadata enrichment ecosystem inspired by the COMET model in collaboration with the stakeholders you've laid out here.

Finally, I’m curious if these enrichments are intended to always come from relatively trusted projects (presumably DataCite was involved or at least reviewed the COMET projects before deploying the enrichments). Will community projects be possible? Will there be mechanisms in place to control who can submit what kinds of enrichments - or to indicate the source of an enrichment? What role might LLM-powered systems play?

That’s correct: DataCite and other community stakeholders have been directly involved in evaluating enrichment projects and outputs produced by COMET. While we are focused on our collaboration with COMET currently, we are conscious of the value of expanding the enrichment pipeline to include the larger community in line with the COMET foundational principles of collective stewardship, de-siloing, and collective benefit.

[tkkuehn]

Thanks for your clarifications and follow-up questions!

Can you describe some of the other ways you might want to find enrichments relevant to you and what you might want to do with those enrichments afterward?

As I alluded to originally, Lunaris depends on RORs in DataCite records to identify Canadian datasets to index in international repositories (e.g. Zenodo). The COMET project to add RORs to DataCite metadata is of great interest to us for this reason! So, if we could query enrichments by the content of the enriched value, that would give us some way to use the enrichments (i.e. by applying them to the full record on our end). An API argument like enriched-value-query with similar capabilities to the dois endpoint's query parameter would be useful, but there are a number of ways this could be presented. Allowing enriched=true for multi-DOI queries would achieve a similar end (with less required post-processing) for us, as you discuss later in your comment.

Being able to filter enrichments by the COMET project with which they're associated would also be nice, maybe using the relationships encoded as resources in some enrichment records, which you described in your comment.

[codycooperross]

Responsive to community feedback, we've continued to iterate on metadata enrichment services:

• Enriched DOI records can now be retrieved in DOI list responses, opening up new ways to explore and retrieve an enriched version of the entire DataCite metadata store using familiar query and filtering tools. Learn more in the August 2026 release notes.
• We’ve added a sourceId attribute to every enrichment record, providing a simple, persistent way to retrieve more comprehensive provenance information about metadata enrichments. The sourceId contains a DOI identifying the COMET project that produced the enrichment record, which can be used to explore project abstracts, related datasets and workflows, connected people and organizations, and more via the DOI metadata and landing page. Learn more in the September 2026 release notes.

We continue to welcome community feedback on metadata enrichment as we work to improve the quality and utility of the DataCite metadata store and enhance discovery, reuse, and reporting across the research ecosystem. Reply to the linked release notes, start a new discussion, or reach out to support@datacite.org for support with metadata enrichment or other DataCite services. Many thanks to all who have contributed already, and we look forward to continuing the discussion!