CORE support for UK HEIs in the REF 2021 Open Access Audit

In June 2019, Research England released REF Audit Guidance which states that CORE data will be used in the REF 2021 Open Access Policy Audit. Following the issuing of the guidance, a number of institutions approached CORE asking what they can do to ensure CORE can accurately collect the required data about their outputs.

The purpose of this document is to provide advice and recommendations to UK HEIs with regards to exposing data to CORE through their repositories. In this document, we use the term repository to refer to a system that the HEI will use for the purposes of REF 2021 as defined in the Deposit requirements, paragraph 235 of the Guidance on submissions.

Contents of this document

1. Key REF 2021 Open Access Policy sections relevant to CORE
2. What the guidance says about the use of CORE in the audit
3. Recommendations for exposing research outputs’ metadata to CORE in support of the audit
4. Why the audit should be supported by an aggregation?
5. How will CORE collect the information?
6. How can CORE help you with making sure your repository is represented well?
7. Subscribe to CORE Repository Edition

For the next Research Excellence Framework (REF 2021), a policy has been introduced for open access. The policy applies to research outputs satisfying the following two criteria:

• Outputs published in a journal or in conference proceedings with an ISSN number.
• Outputs accepted for publication after 1 April 2016.

The policy stipulates that to be eligible for submission to REF 2021, these outputs must be deposited in a repository (institutional, subject, or a repository service) within a specified timeframe. More specifically:

• Outputs accepted for publication from 1 April 2016 to 31 March 2018 must be deposited no later than three months after the date of publication.
• Outputs accepted for publication from 1 April 2018 to 31 December 2020 must be deposited no later than three months after the date of acceptance.

For more details of the policy please see REF Guidance on Submissions, paragraphs 233 to 255.

In June 2019, Research England released REF Audit Guidance which describes how eligibility with the open access policy will be audited. Among other things, the Audit Guidance specifies that CORE data will be used in the audit. More specifically, it states:

40. We will undertake verification of the dates that outputs became publicly available, particularly where they were published early in the REF period or are marked as 'pending' publication (for example, by obtaining a letter from the publisher). This will include checking the publication year against the Crossref database and against Jisc CORE.

46. We will assess each HEIs' overall compliance with the REF 2021 open access policy by… iv. Using Jisc CORE, comparing the datePublished and depositedDate and identifying where the number of days between the two dates is greater than 92.

49. Where there is insufficient evidence to demonstrate a robust and well-managed process for open access, we will identify a set of outputs from each submission made by the HEI, and request further information to verify whether they are compliant with the policy, or whether an exception applies. Outputs may be selected randomly, or based on information in unpaywall.org or Jisc CORE, or a combination of the two. We will select outputs that have been returned as compliant with the policy, and/or outputs that have been returned with exceptions.

The following points list recommendations we make to institutional repositories to ensure their data is well represented in CORE. While these recommendations are not mandated by Research England, implementing them will both improve the quality of the data held by CORE to support the REF Audit as well as increase the discoverability and visibility of the institutions’ research outputs.

1. Make sure that your institutional repository/ies are registered in CORE: We are constantly working towards CORE harvesting all repositories available. However, as repositories sometimes change their web locations, platforms or some institutions introduce multiple repositories, we advise institutions to check that their repositories are registered and listed on the CORE’s data providers page.

2. Adopt RIOXX as a data format: One of the key reasons for the development of the RIOXX metadata application profile was to help institutions achieve compliance with the REF 2021 Open Access Policy. As of June 2019, more than 70 UK HEIs already support RIOXX. The application profile directly supports some of the metadata fields that are important for the audit, such as rioxxterms:publication_date. While CORE can capture the data also from repositories not supporting RIOXX, supporting RIOXX greatly reduces the complexity of the algorithm through which CORE acquires the data. The below recommendations detail which information should be shared in the metadata and how this information should be exposed in RIOXX. We further provide an example RIOXX metadata record with all information needed for the REF Audit correctly set.

3. Add DOIs to records you are planning to submit to REF: The REF 2021 OA policy requires outputs to be deposited within three months of the date of acceptance, which will in many cases mean pre-publication, i.e. at a point when the DOI may not yet be known. To ensure the REF research outputs can be unambiguously matched to the metadata in the repository, we recommend that institutions ensure that DOIs (or other persistent identifiers such as PubMed ID, where appropriate) are added to records to be submitted to REF. This should be done sufficiently in advance of the audit taking place for all outputs where DOI has been assigned. According to RIOXX, DOI of the resource should be put in the rioxxterms:version_of_record field.

4. Release deposit dates publicly, so that they are harvestable by CORE: At present, record deposit dates are not explicitly exposed by repositories in both RIOXX nor standard Dublin Core (DC). We are working with the RIOXX team to encourage explicit sharing of this information in a machine-readable form and will update this guidance once a solution is agreed. In the meantime, and as repositories store the deposit date in their databases, we encourage repositories to expose the deposit date on each of the research output’s web page, even when RIOXX is not implemented, and ideally to mark it as “Date Deposited:”. For example, EPrints repositories typically display the deposit date on the page of a given record (an example at Open Research Online). This date must be the date of the first compliant deposit. We rely on the definition of first compliant deposit.

5. Ensure that dateAccepted field is used where known: The date an output was accepted for publication should be shared in the RIOXX dcterms:dateAccepted field.

6. Ensure that all records to be submitted to REF have a full text linked directly from the dc:identifier: According to the RIOXX standard, the dc:identifier field must contain an HTTP URI which is a link directly pointing to the described resource, i.e. to the MS Word, PDF or another version of the manuscript. This field should be populated for all records your institution is planning to submit to the REF and should appear exactly once in the record, so that the availability of the output in the repository can be automatically checked.

7. Ensure that your repository OAI-PMH endpoint is operational: The OAI-PMH endpoint in your repository enables aggregators to collect data from your repository. The Confederation of Open Access Repositories (COAR) states that the real benefits of repositories come from interoperability which gives us the ability to exchange data within a network. This is currently realised by the OAI-PMH protocol. In the majority of cases, the OAI-PMH endpoint will be correctly working in your repository by default. To check its operation there are freely available validators that ensure the correct response of the endpoint. However, the validator does not fully ensure that the data from the repository could be harvested without issues, which could be a potential problem, especially for larger repositories. Any repository can register freely for the CORE Repository Dashboard account which enables the repository to check its harvesting status.

Below is an example of a RIOXX record with all recommended fields populated:

<?xml version="1.0" encoding="UTF-8"?>
<?xml-stylesheet type="text/xsl" href="/rioxx2.xsl"?>
<rioxx>
  xmlns="http://www.rioxx.net/schema/v2.0/rioxx/"
  xmlns:dc="http://purl.org/dc/elements/1.1/" 
  xmlns:dcterms="http://purl.org/dc/terms/" 
  ...
>
  <dc:title>
    Article title
  </dc:title>
  <rioxxterms:author id="http://orcid.org/0000-1234-5678-9000">
    Author name
  </rioxxterms:author>
  <dc:language>
    en
  </dc:language>
  <dc:publisher>
    Article publisher (In Press)
  </dc:publisher>
  <dc:identifier>
    http://repository.ac.uk/12345.pdf
  <dc:identifier>
  <dcterms:dateAccepted>
    2019-07-31
  </dcterms:dateAccepted>
  <rioxxterms:publication_date>
    2019-09-19
  </rioxxterms:publication_date>
  <rioxxterms:version_of_record>
    http://dx.doi.org/10.1000/abcd.1234.5678
  </rioxxterms:version_of_record>
  ...
</rioxx>

1. Individual HEIs have a local and incomplete view of compliance: An output deposited late to a repository maintained by one institution can still be compliant due to a timely deposit at another repository. Research collaborations are international, i.e. the first compliant deposit can be to an overseas repository and this could also include non-institutional subject repositories, such as arXiv. While a few institutions are involved in cross-institutional checking for possible external deposit, this practice requires human labour and is limited only to cooperating institutions. Aggregators have a global and more complete view over the data and can provide technology that will effectively support institutions in better understanding which outputs are compliant and which are not even prior to the REF submission.

2. Open Access uptake monitoring: A key intention of the REF 2021 OA Policy is to increase the proportion of OA outputs and decrease the time lag between acceptance and deposit. As a significant proportion of web transactions for accessing research papers is to recent papers, if research becomes available earlier in repositories than on publisher platforms and is, in addition, available without a paywall, this will a) significantly increase the importance of the repository infrastructure, b) accelerate research dissemination by making the scholarly communication process more efficient and c) provide better value to citizens. Citing Lord Kelvin's “If you cannot measure it, you can not improve it,” it is crucial, for any policy, that its uptake can be monitored and improvements measured at any time. By capturing this information at an aggregation level, we are enabling precisely that. For example, if institutions were to delivere deposit dates to Research England as a one-off in csv only at the time of REF submission, this would not allow such monitoring, as it would only provide deposit dates for a small subsample of content in repositories and only at one point in time. It is is crucial to realise that the importance of the REF 2021 OA policy goes far beyond REF. The practice of depositing early should prevail beyond the REF submission date in 2020 as we should be able to monitor our journey to not just open access but also to fast open access. In the future, similar policies (e.g. due to Plan S) will require similar monitoring, so it makes sense to get ready now.

3. Audit transparency: Exposing deposit dates and other important information in a machine readable format in support of the audit through repositories ensures transparency of the audited data. Everyone can check the data that not only gets submitted but also that is about to get submitted, ensuring good preparation leading to no surprises. Overall, transparency is a key to good research practices and open science. It also often leads to more attention resulting in improvement of processes and better quality data.

4. Efficiency: Collecting data for the audit through an aggregation is by far the most efficient method of assembling the data. Where systems are already in place, it does not require any manual work or action. Any work likely to be conducted improves the repository infrastructure for beyond REF 2021 resulting in a positive change across the sector. As it is expected that similar audits might need to be carried out in the future, it is sensible to automate this process now.

5. Verifiability: Assembling data at an aggregation level offers verifiability of the declared data. While this hasn’t been implemented so far, it will be possible to test both the presence and machine availability of the full text, which is key to many use cases, including text and data mining. Exposing data, such as deposit dates, will also be seen as a positive development by publishers. From the perspective of CORE, our intention to test these points is motivated purely by our strive to feedback any issues back to repositories and support them in creating a better and more interoperable repository infrastructure, which we all deserve.

CORE regularly harvests data from repositories. If all the recommendations described above are followed, then the procedure is straightforward. If any deviations are present, CORE will try to collect the data in alternative ways. These include:

• If RIOXX is not supported, CORE will gather metadata, such as information about the identifier and DOIs, using standard DC.

• If DOIs are not available for some records, CORE will use available metadata to match these records against several key databases, including Crossref and Microsoft Academic Graph (MAG), to discover the DOIs.

• If the date of publication is not supplied by the repository or it doesn’t have the desired granularity (we prefer ISO 8601 (post–2004 versions) which follows the following format: YYYY-MM-DD), CORE will collect the publication date from Crossref. Where the publication date is supplied by the repository but the date is not the same as in Crossref, the decision as to which one will be used in the audit will be up to the discretion of Research England.

• As deposit dates are not part of standard DC and not yet part of RIOXX, CORE therefore implements the following two methods for collecting them:

  • Where deposit dates are visible through the web pages in the repository (see Item 4 above), CORE attempts to collect them by scraping them, i.e. extracting the data from the web page of the record, using custom built scrapers for a variety of repository platforms.

  • Where deposit dates cannot be obtained using scraping, CORE makes use of the record’s Datestamp provided in the record’s metadata. This is done in the following way. CORE stores the first Datestamp it receives for a given record, i.e. the first time CORE harvests the record, as the deposit date. Subsequent changes to the Datestamp, which could be due to the record’s updates, will not affect the deposit date. Due to the frequent harvesting, this provides a close estimation of the deposit date.

This describes the method we use to collect data from repositories that currently do not support RIOXX. We might update this method if we identify a way of improving it, such as to make the data we collect more accurate.

When on rare occassions the collected data might be incomplete, it will be up to the discretion of Research England perfroming the audit to interpret the data. We believe that if a record present in the repository has not been harvested by CORE or it was not possible to capture some metadata, this should not negatively affect the audit result. To this end, our aim is to support institutions to identify and resolve any issues prior to the audit.

CORE Repository Dashboard already provides useful information to check that your repository is being properly harvested. Additionally, we have released a brand new version of the CORE Repository Dashboard, from which the deposit dates can be retrieved and we will release it soon. We also offer a free CORE Discovery repository plugin which can help repositories discover Open Access full text versions of articles for metadata records in repositories which do not have a full text manuscript attached.

To answer further questions, we have also established and maintain an FAQ section for the purposes of the audit. If you have any further questions, please contact the CORE team for support and we will be happy to help.

The new CORE Repository Edition has now been launched. This new edition contains both new tools for data enrichments and to enable repositories to help ensure compliance with the REF 2021 Open Access Policy.

Visit the CORE Repository Edition service for further details. To subscribe contact us via email at sales@core.ac.uk.