Recent content in Thoughts on Open on Sesame Open Science

Bianca Kramer

On May 23, 2023, the Council of the European Union approved a set of conclusions on


 
 ‘High-quality, transparent, open, trustworthy and equitable scholarly publishing’
 


 (see pdf).


 Below are some of my thoughts on these conclusions.

I think these Council Conclusions are very welcome and important, as they align with, and in turn can support, a number of important current developments in how

Comments on EC Council Conclusions on scholarly publishing

Abstracts are increasingly important given the rise of large language models (LLM) and generative AI. While OpenAlex provides a source of open abstracts in addition to Crossref, the takedown of abstracts from OpenAlex by two major publishers points to an increased commodification of abstracts.

The case for open abstracts Launched in 2020, the Initiative for Open Abstracts (I4OA) advocates and promotes the unrestricted

More open abstracts?

Thoughts on Open on Sesame Open Science

*Abstracts are increasingly important given the rise of large language
models (LLM) and generative AI. While OpenAlex provides a source of open
abstracts in addition to Crossref, the takedown of abstracts from
OpenAlex by two major publishers points to an increased commodification
of abstracts.*

### The case for open abstracts [![](data:image/svg+xml;base64,PHN2ZyBjbGFzcz0iYW5jaG9yLXN5bWJvbCIgYXJpYS1oaWRkZW49InRydWUiIGhlaWdodD0iMjYiIHdpZHRoPSIyNiIgdmlld2JveD0iMCAwIDIyIDIyIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPgogPHBhdGggZD0iTTAgMGgyNHYyNEgweiIgZmlsbD0iY3VycmVudENvbG9yIiAvPgogPHBhdGggZD0iTTMuOSAxMmMwLTEuNzEgMS4zOS0zLjEgMy4xLTMuMWg0VjdIN2MtMi43Ni4wLTUgMi4yNC01IDVzMi4yNCA1IDUgNWg0di0xLjlIN2MtMS43MS4wLTMuMS0xLjM5LTMuMS0zLjF6TTggMTNoOHYtMkg4djJ6bTktNmgtNHYxLjloNGMxLjcxLjAgMy4xIDEuMzkgMy4xIDMuMXMtMS4zOSAzLjEtMy4xIDMuMWgtNFYxN2g0YzIuNzYuMCA1LTIuMjQgNS01cy0yLjI0LTUtNS01eiIgLz4KIDwvc3ZnPg==){.anchor-symbol}](#the-case-for-open-abstracts) {#the-case-for-open-abstracts}

Launched in 2020, the [Initiative for Open Abstracts
(I4OA)](https://i4oa.org/){target="_blank" rel="noopener"} advocates and
promotes the unrestricted availability of abstracts of scholarly output,
particularly journal articles and book chapters, in trusted repositories
where they are open and machine-accessible. This has gained increased
relevance with the rise of large language models (LLM) and generative
AI. Access to high-quality textual content for training purposes is
highly sought after, with the risk of further commodification by
publishers and by developers of generative AI models and tools.

As part of a broader movement towards open research information,
including the [Barcelona Declaration on Open Research
Information](https://barcelona-declaration.org){target="_blank"
rel="noopener"}, I4OA calls on all scholarly publishers to open the
abstracts of their published works and, where possible, to submit them
to Crossref. In October 2024, the proportion of current journal
articles^[1](#fn:1){.footnote-ref role="doc-noteref"}^ published by
Crossref members with open abstracts had reached 52%, up from 21% when
I4OA was launched in September 2020. This is thanks to a [growing number
of
publishers](https://www.crossref.org/blog/i4oa-hall-of-fame-2023-edition/){target="_blank"
rel="noopener"} who deposit abstracts to Crossref, often for close to
100% of their journal articles.

However, there are still large gaps in the availability of abstracts in
Crossref. To some extent, these gaps are known (e.g. where publishers do
not submit any abstracts at all), but to some extent they are also
hidden, since not all journal content necessarily has abstracts (think
editorials, letters to the editor, book reviews etc). Depending on the
discipline and type of journal(s) in its portfolio, a proportion of 70%
open abstracts for a given publisher may in practice mean full coverage,
or may indicate there are additional abstracts that are not (yet) openly
available.

#### OpenAlex as an additional source of open abstracts [](#openalex-as-an-additional-source-of-open-abstracts) {#openalex-as-an-additional-source-of-open-abstracts}

[OpenAlex](https://openalex.org/){target="_blank" rel="noopener"} is
another source where abstracts are brought together in an openly
available and machine-readable format. This provides the opportunity to
estimate the value of OpenAlex in providing additional coverage of
abstracts compared to Crossref. In addition, it can give an indication
of the lower bound of the proportion of journal content from any given
publisher that can be expected to have abstracts (and that publishers
thus can be expected to be able to provide to Crossref).

While abstracts in Crossref are deposited by publishers themselves,
OpenAlex collects abstracts from [external
sources](https://help.openalex.org/how-it-works/entities-overview){target="_blank"
rel="noopener"}. These include Crossref, but also other structured
sources like PubMed, and content obtained by web crawling, including
from publisher landing pages ( [Priem et al.,
2022](https://doi.org/10.48550/arXiv.2205.01833){target="_blank"
rel="noopener"}). In addition, OpenAlex has inherited content previously
collected by Microsoft Academic Graph in similar ways prior to its
cessation in December 2021. Each abstract is included in OpenAlex as an
" [inverted
index](https://docs.openalex.org/api-entities/works/work-object#abstract_inverted_index){target="_blank"
rel="noopener"}", encoding information about the abstract's words and
their positions within the text. OpenAlex content is made available
under a [CC0
license](https://creativecommons.org/publicdomain/zero/1.0/){target="_blank"
rel="noopener"}, allowing unrestricted use and distribution.

#### Comparing open abstracts in OpenAlex and Crossref [](#comparing-open-abstracts-in-openalex-and-crossref) {#comparing-open-abstracts-in-openalex-and-crossref}

Since the launch of I4OA, we have been tracking the openness of
abstracts for all Crossref members over time (for data and code see this
[GitHub repository](https://github.com/bmkramer/I4OA){target="_blank"
rel="noopener"}). For a subset of 40+, mostly larger, publishers, the
proportion of current journal articles that have abstracts deposited in
Crossref is shown in a [chart on the I4OA
website](https://i4oa.org/#:~:text=are%20shown%20in-,orange,-.){target="_blank"
rel="noopener"}, which is updated quarterly.

For this set of publishers, we compared the proportion of journal
articles with abstracts available in Crossref and
OpenAlex^[2](#fn:2){.footnote-ref role="doc-noteref"}^ in October 2024
for publication years 2022-2024 (Figure 1). Only journal articles with
Crossref DOI were included.

![featured image](featured_202411_open_abstracts.png) *Figure 1.
Proportion of journal articles (2022-2024) from selected publishers that
have open abstracts in OpenAlex (dark grey dots) and Crossref (orange
and blue dots for publishers supporting and not (yet) supporting I4OA,
respectively). Publishers are ordered by ascending proportion of open
abstracts in Crossref and descending number of DOIs.*

These data show that for most publishers with less than 90% open
abstracts in Crossref, OpenAlex has additional abstracts, obtained from
additional sources (including through web scraping). This includes many
of the larger publishers that do not (yet) deposit abstracts to Crossref
at all, like Elsevier, IEEE, ACS and Taylor & Francis. Overall, the
proportion of journal articles from 2022-2024 with Crossref DOIs for
which abstracts are available in OpenAlex is 83%, vs 52% with abstracts
available in Crossref. Apart from demonstrating OpenAlex as a valuable
source of machine-readable abstracts for research articles, this also
points to the sizeable potential for improving abstract coverage in
Crossref for many publishers, including those that do already provide
abstracts for part of their journal content.

#### Publishers with limited abstract coverage in OpenAlex [](#publishers-with-limited-abstract-coverage-in-openalex) {#publishers-with-limited-abstract-coverage-in-openalex}

There are some publishers for which the proportion of abstracts
available in OpenAlex is more limited. A general reason for limited
abstract coverage in OpenAlex could be the disciplinary effect mentioned
above - i.e. less journal content that has abstracts in the first place.

This might be a factor especially for publishers that operate
predominantly in the humanities, like Brill, OpenEdition and CAIRN (a
publishing platform in the humanities and social sciences representing a
number of individual publishers). Another example is the American
Psychological Association (APA), where the relatively low proportion of
open abstracts can likely be explained by the inclusion of DOIs for
supplementary materials, most of which do not have abstracts.

In cases where the proportion of abstracts in OpenAlex is essentially
the same as in Crossref, this could represent the maximal potential
(i.e. the remaining content does not have abstracts at all), or,
alternatively, it could mean OpenAlex only takes data from Crossref, and
does not also crawl the publisher domain itself.

#### Takedown of abstracts by Springer Nature and Elsevier [](#takedown-of-abstracts-by-springer-nature-and-elsevier) {#takedown-of-abstracts-by-springer-nature-and-elsevier}

In November 2022, OpenAlex took the step to remove abstracts for
Springer Nature's non-open access content (see [GitHub
commit](https://github.com/ourresearch/openalex-guts/commit/697b7ecad0c377d6381fd7d3b8608e9b917c76ac#diff-9b971b24d1a1af9cbe2c01f368a6408f0d1f69df0f7a0f1f23bc652d7c4f0abaR851){target="_blank"
rel="noopener"}). Springer Nature similarly supplies abstracts to
Crossref only for its open access content ( [van Eck and Waltman,
2024](https://doi.org/10.31222/osf.io/smxe5){target="_blank"
rel="noopener"}). OpenAlex still has more abstracts for Springer Nature
than Crossref (48% vs 34%). This can be explained by two observations:

- Springer Nature only makes abstracts from articles with a CC BY
  license available in Crossref (so excluding CC BY-NC, CC BY-NC-ND and
  other licenses)
- green open access articles (where an open access version of the
  article is available in a repository) continue to have abstracts
  included in OpenAlex.

Now, two years later (in November 2024) Elsevier seems to have taken
similar steps to have abstracts from closed articles removed from
OpenAlex (see [GitHub
commit](https://github.com/ourresearch/openalex-guts/commit/b85b3bc77cf9c0f3bd162426a2ba0dacdc951065){target="_blank"
rel="noopener"}). The effect of this step is already visible in the
OpenAlex web UI and API, and is expected to affect the data snapshot
from November 2024 forward. With previous abstract coverage for Elsevier
at 82% in OpenAlex, this will result in the loss of approximately 1,1M
abstracts for [current closed journal
articles](https://openalex.org/works?page=1&filter=primary_location.source.publisher_lineage%3Ap4310320990,open_access.is_oa%3Afalse,type%3Atypes%2Farticle,primary_location.source.type%3Asource-types%2Fjournal,publication_year%3A2022-2024){target="_blank"
rel="noopener"}, and up to 11.5M abstracts for [closed journal articles
of all
years](https://openalex.org/works?page=1&filter=primary_location.source.publisher_lineage%3Ap4310320990,open_access.is_oa%3Afalse,type%3Atypes%2Farticle,primary_location.source.type%3Asource-types%2Fjournal){target="_blank"
rel="noopener"}.

#### Discussion and conclusions [](#discussion-and-conclusions) {#discussion-and-conclusions}

It is encouraging to see considerable abstract coverage in both Crossref
and OpenAlex, as it increases the corpus of machine-readable abstracts
which are publicly available for use. The fact that OpenAlex in many
cases manages to capture more abstracts than are available in Crossref,
even for publishers who do provide abstracts for some of their articles
to Crossref, points to a gap that publishers themselves can close if
they so choose.

Regarding the use of scholarly material, including abstracts, for the
development of LLM and generative AI applications, this still poses many
questions (see e.g. [Pooley
2024](https://doi.org/10.54900/zg929-e9595){target="_blank"
rel="noopener"}). These include:

- the legality of using licensed or unlicensed copyrighted material,
  both for training and for producing new texts, and the question
  whether such new texts can be considered derivatives;
- the implications of various Creative Commons licenses in this context
  (i.e. can CC-BY licensed material be used in bulk without individual
  attribution?);
- the role of fair use (in the US) and sui generis database rights (in
  the European Union), respectively, to enable and/or restrict use of
  scholarly material;
- the desirability of the use of scholarly material in commercial
  generative AI tools (including these developed and used by
  publishers), as well as the desirability of commercial publishers
  licensing scholarly material to developers of generative AI models and
  tools, as an additional revenue stream;
- the public benefit of using high-quality textual information in the
  development of generative AI applications.

In these discussions, abstracts hold a somewhat special position as they
are treated both as metadata (generally not subject to copyright) and as
part of creative works (which are subject to copyright).

After Springer Nature, Elsevier is now the second publisher that has
taken steps to have abstracts removed from OpenAlex. Coupled with
announcements earlier this year of Elsevier [announcing their own AI
tool](https://www.elsevier.com/products/scopus/scopus-ai){target="_blank"
rel="noopener"} which makes ample use of abstracts, as well as [selling
the right to use and display abstracts from its own journals to a third
party tool
developer](https://www.sspnet.org/community/news/cactus-and-elsevier-join-hands-to-bring-19m-elsevier-research-abstracts-to-researchers-mobile-devices/?utm_campaign=coschedule&utm_source=twitter&utm_medium=ScholarlyPub&utm_content=Cactus%20and%20Elsevier%20Join%20Hands%20to%20Bring%2019M%2B%20Elsevier%20Research%20Abstracts%20to%20Researchers%27%20Mobile%20Devices){target="_blank"
rel="noopener"}, this makes clear that abstracts are in growing danger
of being treated as a commodity, rather than as open research
information.

Open abstracts, either in Crossref or in OpenAlex (where they are made
available with a CC0 public domain dedication, but are subject to
takedown), provide a welcome antidote to this development. Not only does
this make abstracts publicly available for use and re-use, it also
serves as an opportunity for the scholarly community to discuss how open
we want abstracts to be and how to achieve that.

#### Code and data availability [](#code-and-data-availability) {#code-and-data-availability}

- Code:
  [https://github.com/bmkramer/more_open_abstracts](https://github.com/bmkramer/more_open_abstracts){target="_blank"
  rel="noopener"}
- Aggregated data:
  [https://doi.org/10.5281/zenodo.11580550](https://doi.org/10.5281/zenodo.11580550){target="_blank"
  rel="noopener"}

The analysis underlying this work was performed using Curtin Open
Knowledge Initiative (COKI) infrastructure, which is documented on
GitHub:
[https://github.com/The-Academic-Observatory](https://github.com/The-Academic-Observatory){target="_blank"
rel="noopener"}. It uses data snapshots of Crossref and OpenAlex
ingested into a Google Big Query environment where they can be queried
via SQL.

::: {.footnotes role="doc-endnotes"}

------------------------------------------------------------------------

1.  ::: {#fn:1}
    For the purpose of registration and analysis, Crossref defines
    "current" as the current calendar year and the previous two calendar
    years. [↩︎](#fnref:1){.footnote-backref role="doc-backlink"}
    :::

2.  ::: {#fn:2}
    The results reported here only represent the absence/presence of any
    text in the abstract field in OpenAlex and Crossref. This does not,
    in itself, say anything about the quality and completeness of the
    abstract in either database. Previously, concerns have been raised
    about the [quality of abstracts in Microsoft
    Academic](https://www.leidenmadtrics.nl/articles/why-openly-available-abstracts-are-important-overview-of-the-current-state-of-affairs#:~:text=abstracts%20suffer%20from%20some%20quality%20problems){target="_blank"
    rel="noopener"}, and it would be interesting to do a follow-up
    investigation into the quality of abstracts collected by OpenAlex
    (e.g. abstracts for articles published from 2022
    onwards). [↩︎](#fnref:2){.footnote-backref role="doc-backlink"}
    :::
:::


![featured image](featured_202306_ec.png)

*On May 23, 2023, the Council of the European Union approved a set of
conclusions on* ***'High-quality, transparent, open, trustworthy and
equitable scholarly publishing'*** *(see
[pdf](https://data.consilium.europa.eu/doc/document/ST-9616-2023-INIT/en/pdf){target="_blank"
rel="noopener"}).*  
*Below are some of my thoughts on these conclusions.*

I think these Council Conclusions are very welcome and important, as
they align with, and in turn can support, a number of important current
developments in how open access is perceived and implemented. One of
these is the growing awareness, also among research institutions and
funders, of the inequality and unsustainability of APC-driven open
access publishing, especially by commercial or otherwise revenue-driven
publishers. Another is a growing emphasis on important qualitative
aspects of open access, in particular immediacy and open licensing.

In this sense, the Council Conclusions, together with for example the
[UNESCO Open Science
Recommendations](https://en.unesco.org/science-sustainable-future/open-science/recommendation){target="_blank"
rel="noopener"} and the [Action Plan for Diamond Open
Access](https://www.scienceeurope.org/our-resources/action-plan-for-diamond-open-access/){target="_blank"
rel="noopener"}, to name but two, can help shift focus to other models
of publishing. One thing I would like to emphasize is that while
developing new publishing infrastructures (for example at national or
even European level) is in itself a worthwhile avenue to pursue, many
examples also already exist of non-profit publishing solutions that
deserve support and recognition.

As an aside, an important point in the Council Conclusions is that
"costs should not be covered by individual authors or readers" also
means "authors can publish their work without funding/institutional
eligibility criteria". This is especially interesting in light of the
endorsement of Open Research Europe, which at least currently does not
meet this criterium. If the Council Conclusions can be read as an
encouragement to broaden the Open Research Europe model in that sense,
this would be a very positive development in my opinion.

The Council Conclusions are also very explicit in their concern for
increasing costs and their call for transparent pricing commensurate
with publication services. This can support other initiatives in this
area (like cOAlition S Price Transparency Framework) and a more general
reconsideration on how public money is spent. One remark (which I often
make in this context) is that to be truly informative, such price
transparency needs to include transparency on profit or surplus.

Two areas where I would have liked the Council Conclusions to go further
is, first, emphasizing the importance of secondary publishing rights -
as this is something that can be taken up by member states as a
legislative approach (as it has been in a number of European countries)
to enable open sharing of research publications, in particular immediate
sharing. Most existing legislation does not go that far yet, and I feel
the Council Conclusions could have been stronger in pushing this
forward.

Secondly, in the paragraph on monitoring and the "lack of data and
trustworthy information on the state of scholarly publishing" (including
bibliometric data), while mention is made of the need for such data to
be FAIR, I would have liked to see a stronger stance on the importance
of open metadata on scholarly publishing, to enable transparency and
equitable access to such data. In my view, this is even more important
than attempting to ' reduce the fragmentation of monitoring
initiatives' - as different entities will have different needs and
priorities in monitoring, and open availability of good quality metadata
can enable such diversity while still being comparable and transparent.

As a final observation, the Council Conclusions rightly put much
emphasis on the importance of quality, transparency and integrity in
scholarly publishing, and also mention the need for a change in research
culture "that recognises diverse research activities with the
overarching goal to maximise high quality and impact of research", also
referring to the activities of the [Coalition for Advancing Research
Assessment (CoARA)](https://coara.eu/){target="_blank" rel="noopener"}.
In my opinion, these two aspects are related and together could be one
of the keys to the question of costs and pricing - if a change in
research culture would reduce the pressure to publish in so-called
"high-impact" journals for the sake of prestige, and instead focus on
methodological quality, openness aspects and reaching the audiences for
which the research is most relevant (including societal partners), this
would create a more level playing field where "paying for prestige"
could be replaced by paying for actual quality, without selectivity for
selectivity's sake.

While the Council Conclusions do not go that far, I am hopeful that the
recommendations made can help support a transition into that direction.
