Enable Data Citations

[jalperin]

Data Citations

Support for structure Data Citations

This work adds a new Data tab in Publication metadata view which depending on the context metadata settings includes the data availability statement field and/or a data citation manager component. These elements are also visible in submission wizard if so selected.

The Data Citation Manager handles structured Data Citations and these citations are visible in submission landing page.

Current pr's for review:
ojs: pkp/ojs#5204
ops: pkp/ops#1203
omp: pkp/omp#2234
pkp-lib: #12079
ui-library: pkp/ui-library#746

Missing:

• If needed, support for Native import/export
• Support in third party plugins exporting article metadata
• OAI

Data Availability statement (added in 3.5)

Describe the problem you would like to solve

• As a reader of an academic journal article, I sometimes want to see the underlying data that was used in the research.
• As a reviewer of an academic journal article, I sometimes want to see the underlying data that was used in the research.

Describe the solution you'd like
Many journals now require a data availability statement to go along with the published article. These are often collected as part of the original submission, and an opportunity to enter them again/update when uploading a revision.

Who is asking for this feature?
I don't know of specific OJS journals that are requesting this feature, but it has been a standard feature in virtually ALL journal submissions systems (in PLOS since 2014). I imagine there is quiet demand for this, as number of datasets published continues to grow.

Additional information
At a basic level, this information should be collected during the submission/revision process and then published on the article page. This information also typically appears in article PDFs. It can probably also be included in Crossref deposits.

Ideally, we would eventually also support integration with data repositories, such as Harvard Dataverse, Zenodo, and Dryad. Some of these provide private data access links that can be shared with reviewers and editors during the peer review process. Some also have mechanisms for updating the datasets to public once the article is accepted/published.

@dlowenberg from Dryad has lots and lots of information about this topic and has offered to consult. She designed this workflow at PLOS and has seen it in practice at hundreds of journals as part of her work with Dryad.

Related entry: Issue #4176

[NateWr]

Thanks @jalperin. What metadata should be collected that is not currently collected? And what metadata indexing services collect this metadata (and in what form)? Also interested to hear from @dlowenberg if you know.

[dlowenberg]

A couple of things to parse through here: there are data availability statements and there are data citations.

Data Availability Statements

• Statement to appear in the metadata submission form (corresponding author to enter) and displayed on the published article (for readers to be notified of where the data live)
• Free text asking for the identifier for where the data live, e.g., "Data are available at Dryad: https://doi.org/10.5061/dryad.tdz08kpxg"
• They are good for capturing this information and displaying it to contributors but are not standardized or sent to Crossref (note: Crossref is looking into having a space in their metadata for data statements in 2021)

Data Citation

• When identifiers for datasets are included in the article, they should be sent to Crossref as part of the metadata (https://www.crossref.org/education/reference-linking/data-and-software-citation-deposit-guide/)
• You could put in a feature to capture this information from a data statement and include it in the metadata as a relationship type, OR, you can ask authors to include data citations in the reference list as noted in the above Crossref link
• Data Statements are for notifying users how to find the data, Data Citations are for bettering the metadata and indexing data citations so we can build upon these for research purposes and building an incentive/reward system for data

I could send example screenshots of how different journals are currently capturing Data Availability Statements or write up text for what can display but this is a relatively simple metadata field that just asks authors to consider their data and designate where it does or will live. Data Citation work is likely down the road but something you all should consider!

Let me know what else would be helpful.

[NateWr]

Thanks @dlowenberg, that's really helpful! Just to get something clear in my head: when you talk about data citations, are you talking about data sets cited within the research article itself? So these data sets may or may not be from the author(s)? In writing my article, I may, for example, cite a data set that has been put up somewhere by a city council or something, and later cite a data set that our own research has produced and uploaded to an instance of Dryad or Dataverse?

I'm trying to tease out the difference -- if there is one -- between identifying a "relationship" (eg - this DOI is a data set produced by/created for that DOI) vs. a "citation" (eg - this DOI cites that DOI). Is that a sensible distinction to draw? Should it be reflected differently in the metadata we produce?

I could send example screenshots of how different journals are currently capturing Data Availability Statements

That would be helpful! I think we have an idea of what this would be, but examples are good at quickly identifying misunderstandings.

[dlowenberg]

Data citations can be self citations or citations to other datasets that were used. Your two examples would both qualify. That is why there are two methods for capturing these and sending to Crossref, one being a relation type like a self citation saying "this is my dataset related to this article" and the other being in a reference list referencing another public dataset.

In terms of how this is really a "citation" - that is something we are all working on from Make Data Count, understanding how datasets are "cited", what it means, etc. So in terms of technical implementation, no need to think about those distinctions :)

[asmecher]

(Just a quick note that we need to consider this as much for reviewers as for authors. Reviewer access to data is a frequent subject around here these days.)

[heike-r]

As journal editor, I would greatly appreciate an additional metadata field in the article metadata. Some of our authors deposit data in specific repositories (DNA sequences, automatically generated images, satellite data, spectroscopic data, raw microscopy images, ....) and we would like to ask more of them to do this. However, we would not publish the datasets themselves as there are far to many different datatypes. Thus, we would not need metadata for datasets. What we would appreciate though, is one (or more) additional fields in the article metadata where the author can enter IDs for datasets they uploaded elsewhere. We register article DOIs via DataCite and they have an ambiguous field 'relatedidentifier' in their xml schema. In most cases, this would be a DOI of a dataset. It would be nice if an author can enter related identifiers during the submission, the editor and reviewer have access to them, they are published alongside the article DOI on the article website and are included in article metadata for DOI registration as 'relatedidentifier'.

[NateWr]

Thanks @heike-r! Do you know if DataCite's relatedidentifier specifies what kind of relationship there is? There are a lot of potential things that a related DOI might represent, and I suspect other metadata depositing services might have their own categories. It would be interesting to know what Crossref supports.

If there's enough overlap, it could be a distinct metadata setting. But if the depositing services all define and support different kinds of relationships, it may make sense to make the metadata a part of the depositing plugin (ie - DataCite or Crossref).

[heike-r]

@NateWr
What I gathered from the documentation of the DataCite Metadata Schema 4.3 (https://support.datacite.org/docs/schema-optional-properties-v43) is this:

relatedIdentifier
Definition: Identifiers of related resources. These must be globally unique identifiers.
values: freetext

relatedIdentifierType (mandatory if relatedidentifier is given)
values: controlled list with 19 options (most popular identifiers like DOI, ISSN, URL, w3ID, arXiv, …)

relationType (mandatory if relatedidentifier is given) is the description of the relationship of the resource being registered (A) and the related resource (B).
values: controlled list with 32 options, recommended values are: IsCitedBy, Cites, IsSupplementTo, IsSupplementedBy, IsPartOf, HasPart, IsVersionOf, …

edit:
The CrossRef documentation says, that CrossRef and DataCite use the same kind of linking and the same metadata tags.
https://www.crossref.org/education/content-registration/structural-metadata/relationships/

[NateWr]

Thanks @heike-r, that's helpful! I wasn't sure looking at the DataCite documentation whether there was a relationType that is appropriate for datasets, but the Crossref link you posted specifically mentions datasets.

So for DataCite we'd use the following:

<relatedIdentifier relatedIdentifierType="DOI" relationType=”IsSupplementedBy”>10.1234/...</relatedIdentifier>

There's an alternate relatedIdentiferType of URL that we could use when a DOI does not exist. So it could be:

<relatedIdentifier relatedIdentifierType="URL" relationType=”IsSupplementedBy”>https://...</relatedIdentifier>

Since Crossref uses the same schema (:pray:) that means that we'd need to support one or many text fields for entering related datasets. We can then inspect the entry to determine if it is a URL or DOI, and pass the correct value to DataCite.

cc @asmecher this overlaps somewhat with your discussion of URL-based submission files, but tackles the problem from a different direction.

[bozana]

We now (coming with 3.4 release) have an input field for the data availability statement, s. #8290.
But, I am wondering what would be a good way to capture data citations. @dlowenberg, how do other publisher/publishing systems (e.g. PLOS) do that? (I was looking at PLOS articles, but could not quickly find an article listing them visibly/separately...)

We do provide (normal) citations to Crossref (using the CrossrefReferenceLinking plugin), but according to Crossref documentation it would be better to separate data citations.
(Currently I am thinking if we should have a separate text area input field that will be only for data citations, that would then be each extracted and send to Crossref. They would need to have a URL or DOI i.e. persistent identifier.)