Digital resources in the Social Sciences and Humanities OpenEdition Our platforms OpenEdition Books OpenEdition Journals Hypotheses Calenda Libraries OpenEdition Freemium Follow us

A new revolutionary workflow for a unified collection of citations: say hello to the OpenCitations Index

Blog post by Ivan Heibi (University of Bologna), Arianna Moretti (University of Bologna) and Chiara Di Giambattista (University of Bologna).

In the past five years, the OpenCitations data has been enriched with numerous new indexes of open citation data from different sources. However, the quantity and diversification of the ingested information have raised several issues, which recently made it essential to conduct a complete revision of the ingestion workflow. The result was a revolution in the way OpenCitations data is delivered. In this blog post, we will explain the context and challenges raised by the old procedure. Then, we will present the new ingestion workflow, designed to produce just two comprehensive collections: OpenCitations Index, collecting open citation data, and OpenCitations Meta, for the open bibliographical metadata. 

Once upon a time, there were five OpenCitations indexes…

In 2018, OpenCitations released the kickoff version of its first citation index, COCI (citations from Crossref), which contained around 300 million citation links derived from the subset of the reference lists in the Crossref database, where citing and cited entities were identified using Digital Object Identifiers (DOIs). COCI gathered citations with associated metadata in compliance with the recommendations from the Initiative for Open Citations (I4OC) that citation data should be structured, separable, and open, thus marking a turning point by providing a disruptive and free and open alternative to earlier sources such as Google Scholar, which provided freely accessible data although not downloadable, and Web of Science or Scopus, which demanded paid access. 

In a short time, COCI became a competitive and trusted index of citation data, used by numerous institutional repositories, including B!son and Optimeta. In 2021, COCI was taken into account in a comparative study with the most relevant sources in the landscape, including the proprietary ones, which showed its coverage approaching parity with those of the other sources involved in the analysis (Microsoft Academic, Scopus, Dimensions, and Web of Science). At the time of its most recent update in January 2023, COCI counted more than 1.4 billion citations. The reason behind this outstanding number lies in several factors, including Elsevier’s endorsement of the Declaration on Research Assessment (DORA) in December 2020, leading to the open release via Crossref of the reference lists of the articles published in all its journals, and confirming the value of initiatives such as the Initiative for Open Citations (I4OC)

However, before this change of heart, in 2019 OpenCitations had tried to narrow the open citations coverage gap by launching its second index, the Crowdsourced Open Citations Index (CROCI). This index allowed publishers and scholars to contribute directly by uploading crowdsourced open citations into the OpenCitations infrastructure.

In December 2022, a new concrete step towards a factual plurality of OpenCitations indexes was taken by the ingestion of new data sources into the infrastructure, with the publication of the inaugural dumps of DOCI (citations from DataCite) and POCI (citations from PubMed). In June 2023, the first version of the OROCI (citations from OpenAIRE) dump was released too, and JOCI (citations from JALC) is expected to be available by the end of November 2023, for a total of five collections from different sources. 

Why a new workflow? The issues with multiple sources management and new challenges

While having such a variety and richness of indexes helped present the extent of OpenCitations sources, the recent increment in the number of sources and the diversification of data integrated led to two primary issues:

    1. the necessity to handle the ingestion of new identifier types in a DOI-based software infrastructure, and
    2. the consequent possibility of encountering the same citation expressed by several sources with different identifiers.

Moreover, it soon became evident the need to optimize the reuse of the already developed software components to facilitate the metadata crosswalk processes between the new sources’ data models and the OpenCitations Data Model, with the aim to define a functional and easily extendable workflow to be easily reused when it comes to incorporating new data sources, which should be: 

    1. sufficiently generic to establish a globally unique procedure; 
    2. customizable enough to capture the necessary information within each of the specific data models and formats. 

As a solution, we decided to use OpenCitations Meta, the new OpenCitations database and tool for managing bibliographic data related to the publications involved in the citations. OpenCitations Meta makes it possible to assign each entity involved in a citation an internal identifier, nominally the OpenCitations Meta Identifier (OMID), to which all the associated persistent identifiers of the same publication are redirected.

As a result, the allocation of an OMID for each bibliographic resource also enabled the unambiguous identification of each citation, regardless of the persistent identifier schema originally used by the data source to identify the resources. This approach allowed us to perform data deduplication and finally make all the sources’ contributions converge into a unified index containing all the unique citations managed by OpenCitations, expressed as OMID to OMID citation links.

The revised workflow

The new workflow is based on three main components with the benefit of optimizing the process both in terms of computational cost and in terms of flexibility. As shown in Fig. 1, in a preliminary step, source-specific software converts the input dataset – structured according to the source data model – to extract two OpenCitations Data Model compliant data collections in tabular format for bibliographic metadata and citation data, respectively.

The following steps are common to the process of each dataset.  

STEP 1: The bibliographic metadata collection is used as input for the META software. At this stage, it is checked whether or not the bibliographic entities have been previously integrated into our infrastructure (coming from other data sources). If so, the existing OMID is linked also to the new alternative identifiers of the new bibliographic resources. New metadata values, if any, are also integrated. A new OMID identifier is produced for entities never previously encountered, uniquely representing the bibliographic resource in OpenCitations. The outputs of the process are: (I) an updated version of the OpenCitations Meta collection that also includes the metadata of the bibliographic entities provided by the new source, and (II) a collection of provenance data. An internal database is constantly refreshed to preserve correspondence between IDs and the associated internal OMIDs.

STEP 2: Starting from the collection of citations expressed as directional links between identifiers of potentially any type (e.g., DOI-DOI, PMID-PMID, PMC-PMID, etc.), the INDEX software queries the internal database mapping IDs to OMIDs to produce an updated version of the OpenCitations Index: unique citations expressed as OMID-OMID links in different formats, accompanied by their corresponding provenance data.

Fig. 1: An overview of the data ingestion workflow, starting from the data source-specific conversion and production of citations and bibliographic metadata tables, progressing through the META process and the assignation of an OMID identifier to each bibliographic record involved in a citation, and culminating with the exposition of the OpenCitations Index collection of OMID-OMID unique citations.

What we have now: The OpenCitations Index 

From now on, OpenCitations will no longer display an index of citation data for each source. Instead, we will publish a single collection of citations into which the contributions from each of the sources will flow, which we will simply call ‘The OpenCitations Index‘. The first version of this unified index of OMID-OMID citations is posted on Figshare. It was produced in RDF, CSV, and SCHOLIX formats, together with a collection of its provenance information, provided in RDF and CSV formats. For each citation, it is possible to trace the source of the information by consulting the Provenance data collection, thanks to the http://www.w3.org/ns/prov#atLocation property, which defines the location of each citation.

This new solution has the benefit of simplifying the consultation of the data maintained by our infrastructure without reducing the information content. In addition, by including efficient handling of the deduplication problem, the new Index not only provides accurate data on the exact number of unique citations exposed by the framework but also verifies the individual contribution of each source, as well as their overlapping data (Fig. 2).

Fig. 2: An overview of the number of citations stored in the OpenCitations Index as of October 31, 2023. The diagonal cells in the table (highlighted in yellow) show the unique contribution of each collection to the OpenCitations Index, while the other cells represent the citations that are shared between the collections. More in detail, the green cells show the overall input of each source, while the pink cells represent the number of overlapping citations between two data sources.

Currently, the Index contains almost 2 billion unique citations. By the end of November, a new version of the collection will be published, including the contribution of the new Japan Link Centre (JaLC) source. 

How to access the OpenCitations Index data

To maximize the reuse of the exposed information and to ensure the greatest possible interoperability, the collection will always be published on Figshare in all formats listed above. In addition, the data will be accessible via an API, a SPARQL endpoint, and a web interface.

The redesign of the ingestion workflow marks a fundamental step for OpenCitations towards a more intuitive and simple access to our services while always preserving and improving the quality of our data. If you need further information on how the new workflow works, please visit our website, contact us at contact@opencitations.net  or leave feedback and/or suggestions in the dedicated card on our public roadmap to help us improve our services and communications. Thank you!

Discover POCI, the index of open citations from PubMed 

We’re happy to announce POCI, the OpenCitations Index of PubMed open PMID-to-PMID citations, an RDF dataset containing details of all the citations from publications bearing PubMed Identifiers (PMIDs) to other PMID-identified publications, harvested from the National Institutes of Health Open Citations Collection (NIH-OCC). The citations available in POCI are treated as first-class data entities, with accompanying properties including the citations timespan, modelled according to the OpenCitations Data Model. 

Currently, POCI’s December 2022 release contains 717,654,703 citations from 26,024,862 bibliographic resources, and is based on the dump of NIH Open Citation Collection dated November 2022. 

Citation URLs

Each citation (i.e. an individual of the class cito:Citation) is identified by an URL structured as follows:

https://w3id.org/oc/index/poci/ci/[[OCI]].

Open Citation Identifiers

Each Open Citation Identifier [[OCI]] has a simple structure: the lower-case letters “oci” followed by a colon, followed by two numbers separated by a dash (e.g. https://w3id.org/oc/index/poci/ci/01600102060800080706-016002060909030401), in which the first number identifies the citing work and the second number identifies the cited work.

For citations in which the citing and cited works are identified by PMIDs, which includes all the POCI citations, the OCI is created in the following manner, as explained more fully here. Each converted numeral part of OCI is prefixed by a 0160, which indicates that NIH is the supplier of the original metadata of the citation (as indicated at http://opencitations.net/oci).

OCIs can be resolved using the OpenCitations OCI Resolution Service.

Access to POCI data

All the data in POCI:

What is an Open Citation Index?

A citation index is a bibliographic index recording citations between publications, allowing the user to establish which later documents cite earlier documents. The current indexes available in OpenCitations are:  

All the OpenCitations Indexes have six characteristics in common, summarized here: https://opencitations.net/index   

 

Discover DOCI, the index of open citations from DataCite

We’re excited to introduce DOCI, the OpenCitations Index of Datacite open DOI-to-DOI citations, a new tool containing citations derived from publications bearing DataCite DOIs to other DOI-identified publications, harvested from DataCite. The citations available in DOCI are treated as first-class data entities, with accompanying properties including the citations timespan, modelled according to the OpenCitations Data Model

Currently, DOCI’s December 2022 release contains 169,822,752 citations from 1,753,860  citing resources, and is based on the last dump of DataCite dated 22 October 2021 provided by the Internet Archive

Citation URLs

Each citation (i.e. an individual of the class cito:Citation) is identified by an URL structured as follows:

 https://w3id.org/oc/index/doci/ci/[[OCI]].

Open Citation Identifiers

Each Open Citation Identifier [[OCI]] has a simple structure: the lower-case letters “oci” followed by a colon, followed by two numbers separated by a dash (e.g. https://opencitations.net/index/doci/ci/080010504060836132137200707121027-080010504060836161221130313.html), in which the first number identifies the citing work and the second number identifies the cited work.

For citations in which the citing and cited works are identified by DOIs, which includes all the DOCI citations, the OCI is created in the following manner, as explained more fully here. Each case-insensitive DOI is first normalized to lower case letters. Then, after omitting the initial doi:10. prefix, the alphanumeric string of the DOI is converted reversibly to a pure numerical string using the simple two-numeral lookup table for numerals, lower case letters and other characters presented at https://github.com/opencitations/oci/blob/master/lookup.csv. Finally, each converted numeral is prefixes by a 080, which indicates that DataCite is the supplier of the original metadata of the citation (as indicated at http://opencitations.net/oci).

OCIs can be resolved using the OpenCitations OCI Resolution Service.

Access to DOCI data

All the data in DOCI:

More information is available at https://opencitations.net/index/doci. 

What is an Open Citation Index? 

A citation index is a bibliographic index recording citations between publications, allowing the user to establish which later documents cite earlier documents. The current indexes available in OpenCitations are: 

All the OpenCitations Indexes have six characteristics in common, summarized here: https://opencitations.net/index  

Follow OpenCitations on Mastodon

OpenCitations has happily joined the open-source social media platform joinmastodon.org.

Mastodon is “a free and open-source software developed by a non-profit organization”, with the aim of favouring interoperability and bringing social media interaction “back in the hands of the people”.

We look forward to recreating there our wide network of connections, and getting in touch with new people, projects and institutions in a different virtual environment.

Follow us at https://scicomm.xyz/@opencitations !

Tutorial: how to process COCI’s zipped CSV dump without decompressing it

Blog post by Ivan Heibi (Universiy of Bologna) and Arcangelo Massari (University of Bologna).

OpenCitations publishes the COCI dataset after each new release in three main formats: CSV, N-Triples, and Scholix (see https://opencitations.net/download#coci). The CSV format is the most popular and downloaded one due to its comprehensive data organization (i.e. tabular format) and smaller size (compared to the other formats provided). Therefore, this is also the format we suggest using for a local process of the entire COCI dataset. 

The CSV dumps of COCI are uploaded on Figshare. You can check and download the last dump released from https://doi.org/10.6084/m9.figshare.6741422. The dump consists of one main ZIP file, including other smaller ZIP archives (one for each release) containing the actual CSV files (Figure 1).

Figure 1. The contents of the COCI CSV dataset (after the August 2022 release)

It is possible to process this data without unzipping the internal archives, thus saving a lot of disk space. In this tutorial, we will see how to achieve this in Python. Same process could be done in other programming languages.

Processing the COCI dump using Python

Step 1) Downloading the COCI dump

First, you need to download the last CSV dump release of COCI from https://doi.org/10.6084/m9.figshare.6741422 and decompress only the external archive. After this operation, you should have a folder containing the internal ZIP files such as in Figure 1.

Note: It is beneficial to decompress the external archive because doing so does not increase the space occupied on the disk (compressing archives results in a compression rate of 0%) and because working on nested archives would significantly increase RAM requirements. 

Step 2) Working with the ZIP files

Python provides the built-in zipfile module, whose ZipFile class allows you to create, read, write, edit and list the contents of a ZIP file. Given as input the path of the root directory containing all the ZIP files (FOLDER_PATH), the process elaborates each of these files on a different iteration. Each cycle initializes a ZipFile object by specifying the path to the ZIP file (archive_path).

from zipfile import ZipFile
import os

for archive_name in os.listdir(FOLDER_PATH):        
archive_path = os.path.join(FOLDER_PATH, archive_name)
with ZipFile(archive_path) as archive:     # ...

Step 3) Accessing the ZIP files

Use the namelist() method to return the list of CSV files contained in each archive. Then to open the inner CSV files, simply cycle through the list of names and feed them to the open() method of the ZipFile instance, i.e. archive in the example below.

from zipfile import ZipFile
import os

for archive_name in os.listdir(FOLDER_PATH):
    archive_path = os.path.join(FOLDER_PATH, archive_name)
    with ZipFile(archive_path) as archive:
        for csv_name in archive.namelist():
with archive.open(csv_name) as csv_file:       # ...

Step 4) Reading the CSVs

The .open() method returns a buffer. To read the CSV file as a list of dictionaries (i.e. represent each row of the CSV in dictionary format, e.g., {“column1″:”val1”, “column2″:”val2”}) we need to transform the buffer using the TextIOWrapper class and read it using the DictReader class of csv. Then we convert the result of DictReader into a list. 

from io import TextIOWrapper
from zipfile import ZipFile
import os

for archive_name in os.listdir(FOLDER_PATH):
    archive_path = os.path.join(FOLDER_PATH, archive_name)
    with ZipFile(archive_path) as archive:
        for csv_name in archive.namelist():
with archive.open(csv_name) as csv_file:
reader = csv.DictReader(io.TextIOWrapper(csv_file))
rows = list(reader)
# ...

Step 5) Processing the CSVs content

Now you can go through each row of the list and process the citation data as you want. The following example prints the citing and cited entity of each citation in the dump. 

from io import TextIOWrapper
from zipfile import ZipFile
import os

for archive_name in os.listdir(FOLDER_PATH):
    archive_path = os.path.join(FOLDER_PATH, archive_name)
    with ZipFile(archive_path) as archive:
        for csv_name in archive.namelist():
with archive.open(csv_name) as csv_file:
reader = csv.DictReader(io.TextIOWrapper(csv_file))
rows = list(reader)
# Process the CSV here
for r in rows:
print("Citing entity:",r["citing"])
print("Cited entity:",r["cited"])

 

OpenCitations Access Tokens: how they work and why they are important

Since its inauguration in 2010, OpenCitations has always granted free access to its services to users throughout the world, with no requirement for registration or sign-up. Programmatic access to OpenCitations data can be obtained either via our SPARQL endpoints and our REST APIs. In addition, OpenCitations data – available in CSV, Scholix, and RDF formats – can be downloaded from data dumps made periodically and stored on Figshare, so as to enable large-scale analyses using the whole content of the data sets, and also be obtained via our user-friendly text-based search and browsing interfaces.

One of OpenCitations’ priorities is (and will always be) to keep its data globally open and available at zero cost and without restriction for third-party analysis and re-use. As a matter of sustainability, OpenCitations relies on financial support from the scholarly community, which includes those institutions that use OpenCitations data. However, OpenCitations has not so far had in place a proper system to monitor its users, and the main evidence of the impact of OpenCitations in different academic fields and countries has been incompletely obtained from direct contacts with our members and donors across the world, our collaborations with international projects, and the interactions on our social platforms (Twitter and LinkedIn).

We would now like to institute a system that enables us to follow the usage and assess the impact of OpenCitations more reliably. For this purpose, we are now happy to announce the launch of the OpenCitations Access Token System for access to the OpenCitations data and services.

An OpenCitations Access Token is an opaque character string that anonymously identifies a unique user of the OpenCitations APIs. OpenCitations assigns an access token only if authorized to do so by each user, who can request a token by inserting his/her email address into the access form and clicking “Get token”. Upon submission of such a request, each user will automatically receive a personal access token by email. Users can save their personal access token and reuse it every time calling the APIs of OpenCitations, by passing it as a value for the key access-token in the header of each API call. 

Obtaining and using an OpenCitations Access Token is thus easy. It only requires a simple form request, and then the insertion of your personal token into the API call header when using OpenCitations REST APIs. OpenCitations will not store users’ email addresses or any personal information, so that the users’ privacy will be totally safeguarded. The token system just provides a simple mechanism for identifying unique users, for which the use of IP addresses is insufficient.

Obtaining an OpenCitations Access Token will take the user only a few seconds and needs to happen only once. You can request your OpenCitations Access Token here

https://opencitations.net/accesstoken 

Use of an OpenCitations Access Token is not compulsory. However, token use will help OpenCitations incredibly, by enabling us to monitor the number of the unique users accessing our data and services, providing objective anonymized evidence of the number of institutions and researchers accessing our data either occasionally or on a regular basis, which we can then employ to demonstrate the usefulness of OpenCitations in the research environment. While the token system will initially be employed just for API calls (the most used service we offer), it will subsequently be extended to our other forms of data access.

OpenCitations exists for the people that use its data for research purposes every day, and thanks to their support. This is why obtaining precise knowledge of how many researchers and institutions are accessing our services is essential to us, since it will enable us to present the uniqueness and value of OpenCitations to new communities of stakeholders, and thus to make it possible to enlarge the already enthusiastic and diverse group of people and institutions supporting and using our Open Science Infrastructure.

To summarize: Getting and using an OpenCitations Access Token is voluntary, easy, and does not cost you anything. However, it will help OpenCitations a great deal. Please get your own token now, and use it next time you access OpenCitations. Thank you very much!

 

Additional 48 million citations in COCI, including references from IEEE 

We announce the August 2022 release of COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations, which is based on open references to works with DOIs within the Crossref dump dated August 2022. This new release extends COCI with more than 48 million additional citations, giving a total number of more than 1.36 billion DOI-to-DOI citation links. 

This release includes citations from the articles published over the last four years by IEEE, whose bibliographic references were opened in June 2022. 

A fundamental role in pushing the commercial publishers to open their citation data was played by Crossref’s recent announcement to change its reference distribution policy, by making all its metadata open.  

Besides IEEE, COCI already includes the citation data derived from Elsevier (open via Crossref since December 2020) and from the last articles published by the American Chemical Society (whose references were opened in February 2021) 

You can find more information about COCI in our open-access article  

Ivan Heibi, Silvio Peroni & David Shotton (2019). Software review: COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations. Scientometrics, 121 (2): 1213-1228. DOI: https://doi.org/10.1007/s11192-019-03217-6    

Finally, just a reminder that the bibliographic and citation data in COCI:  

    • can be queried using the OpenCitations Indexes SPARQL endpoint;  
    • can be retrieved by using the COCI REST API;  
    • can be searched by using the OpenCitations Indexes Search Interface;  
    • are also available as dumps on Figshare in CSV, N-Triples, and Scholix; and  
    • can be freely re-used for any purpose.

New documents that present OpenCitations’ mission, unique benefits, present status and future plans

Posted on August 10th 2022 by Chiara Di Giambattista

More than a year ago, Ginny Hendricks, Director of Member & Community Outreach for Crossref, and a valued member of the OpenCitations International Advisory Board, published on the Crossref blog the post “The road ahead: our strategy through 2025”. In order to describe all Crossref’s principles and activities, Ginny presented the Crossref strategic planning framework as a diagram summarizing Crossref’s statements, key messages and truths. The clarity and immediacy of the diagram were such that we adapted it to present  OpenCitations’ own statements and goals. The resulting poster “OpenCitations – what does the future hold?” was presented by our Director David Shotton at the OASPA2021 conference, and can be found in this blog post.

Although the poster offered a wide overview of OpenCitations values, unique traits, benefits and plans, it differed slightly from Ginny’s original diagram, in particular because it lacked a “Mission Statement”, scattering the relevant information within the “Values” and “Principles” boxes. Indeed, at that time (September 2021), we didn’t have a clearly defined Mission Statement.

Nevertheless, the creation of that poster was crucial in helping us start to articulate more clearly the purpose and meaning of OpenCitations. As David underlined in his post “From little acorns…a retrospective on OpenCitations”, since 2018 OpenCitations activities have progressively increased and, with them, the number of related journal articles, conference papers and technical definitions. OpenCitations’ involvement in international networks and collaborations (such as SCOSS and the OpenAIRE-Nexus project), together with our need of identifying and reaching out to new stakeholders to assure OpenCitations’ development and sustainability, has made it necessary to publicly define OpenCitations’ mission, unique strengths and next developmental steps.

After numerous revisions, aided by wise advice from members of the OpenCitations Advisory Board members, we’re now happy to publish the following three OpenCitations documents:

OpenCitations Mission Statement,

The Uniqueness of OpenCitations   and

OpenCitations – Present Status and Future Plans,

which together provide a summary of why we exist and where we are heading.

We are particularly proud of the definition of OpenCitations’ primary mission, namely

to harvest and openly publish accurate and comprehensive metadata describing the world’s academic publications and the scholarly citations that link them, and to preserve ongoing access to this information by secure archiving.

The Mission Statement also presents brief descriptions of the OpenCitations context, our vision, our value proposition and our relationship with the community and stakeholders.

The Uniqueness of OpenCitations provides the answer to the question ‘Why choose to use OpenCitations?’, and is a detailed presentation of OpenCitations’ benefits.

OpenCitations – Present Status and Future Plans summarizes OpenCitations’ ongoing activities, that can be quickly visualized on our public roadmap. It also introduces the OpenCitations Working Groups, served by the members of the OpenCitations International Advisory Board, which are currently working on the themes of governance evolution and community building, with the common purpose of driving OpenCitations along the path from being a ‘sustainable infrastructure’ (in POSI terms) to being an enduring community led and financially sustained infrastructure.

In fulfilling our mission and reaching our goals, the support and vital interest of our community members is fundamental. We request that you, as a member of our community, provide us with feedback on these documents and the ideas they contain, or indeed to ask for clarifications, to help us improving our mission and our communications to explain it. You can reach us here: contact@opencitations.net.

Thank you!

Performing live time-traversal queries on RDF datasets

Guest post by Arcangelo Massari, University of Bologna

In this post, Arcangelo Massari, who recently graduated in Digital Humanities and Digital Knowledge under Professor Silvio Peroni at the University of Bologna, shares the results of his master thesis.

A particular problem in information retrieval is that of obtaining data from an evolving dataset, independent of the time at which that item of data was added, changed or removed. To permit such time-independent queries to be performed over evolving RDF datasets, I have developed two new pieces of open source software, time-agnostic-library [1] and time-agnostic-browser [2], that are now available from the OpenCitations GitHub repository.

The time-agnostic-library is a Python library to perform live time-traversal queries on RDF datasets. Time-traversal means being agnostic about time: a SPARQL query that is not run on the current state of the collection but over its entire history or over a specified timespan of that history [3]. This tool allows materializations – obtaining all versions of an entity over time, or its status at a given time. Furthermore, SPARQL queries can be performed to get the delta between two or more versions of one or more resources. Thereby, the time-agnostic-library realizes all the retrieval functionalities described in the taxonomy by Fernández et al. [3].

To complement this query software, the time-agnostic-browser is a web application built on top of the time-agnostic-library to achieve the same results via a graphical user interface.

The primary purpose of these developments is to offer a system for browsing the provenance [4] of RDF statements across time: who produced them, when, where the information was taken from, and what changes were made compared to the previous state of the resource. Knowledge of such information is essential because data changes over time, either because of the natural evolution of concepts or due to the correction of mistakes. Indeed, the latest version of knowledge may not be the most accurate. Such phenomena are particularly tangible in the Web of Data, as highlighted in a study by the Dynamic Linked Data Observatory, which noted the modification of about 38% of the nearly 90,000 RDF documents monitored for 29 weeks, and the permanent disappearance of 5% of them [5] (Figure 1).

Figure 1. Donut chart showing the results of the study conducted by the Dynamic Linked Data Observatory on the evolution of RDF documents [5].

Additionally, the truthfulness of data cannot be assessed without provenance records and a system to query them. In fact, the truth value of an assertion on the Web is never absolute, as demonstrated by Wikipedia, which in its official policy on the subject states: “The threshold for inclusion in Wikipedia is verifiability, not truth.” [6]. The Semantic Web does not alter that condition, and trustworthiness has to be evaluated by each application by probing the context of the statements [7]. It is a challenging task and thus, in the Semantic Web Stack, trust is the highest and most complex level to satisfy, subsuming all the previous ones (Figure 2).

Figure 2.The Semantic Web layers [7]. Trust is the uppermost level of the stack, subsuming all the others.

Notwithstanding these premises, at present the most extensive RDF datasets – DBPedia [8], Wikidata [9], Yago [10], and the Dynamic Linked Data Observatory [11] – do not use RDF to track changes and record the provenance of such changes. Instead, they all adopt backup-based archiving policies. Some of them, such as Yago 4, record provenance but not changes. As far as citation databases are concerned, OpenCitations is the only infrastructure to implement change-tracking mechanisms and to record full RDF provenance records for each data entity. Among the leading players in this field, neither Web of Science nor Scopus have adopted similar solutions.

In accordance with the OpenCitations Data Model (OCDM) [12], a provenance snapshot is generated by OpenCitations every time a bibliographical entity is created or modified. Each snapshot (prov:Entity) records the responsible agent (prov:wasAttributedTo), the generation time (prov:generatedAtTime), the invalidation time (prov:invalidatedAtTime), the primary source (prov:hadPrimarySource), and a link to the previous snapshot (prov:wasDerivedFrom), using terms from the Provenance Ontology. In addition, OCDM introduced a system to simplify restoring an entity’s status at a given time, by saving the delta between two versions as a SPARQL update query (prov:hasUpdateQuery) [13] (Figure 3). This approach enables one to restore an entity to a specific timepoint (snapshot) in a straightforward way by applying the inverse operations, i.e., deletions instead of additions, etc.

Figure 3. Provenance in the OpenCitations Data Model.

This solution is concretely used in all the datasets related to the OpenCitations infrastructure, such as COCI, an open index containing almost 1.2 billion DOI-to-DOI citation links derived from the open reference data available in Crossref [14]. It is important to note that this OpenCitations provenance model is generic and reusable in any other context. Since the time-agnostic-library leverages OCDM, it too is generic and can be used for any RDF dataset that tracks changes and provenance as OpenCitations does.

The time-agnostic-library is released under the ISC license and is downloadable through pip [1]. Test-driven development was adopted as a software development process during its creation [15]. It makes three main classes available to the user: AgnosticEntity, VersionQuery, and DeltaQuery, for materializations, version queries, and delta queries, respectively (Listing 1).

Listing 1. Code template to achieve materializations, time-traversal queries, and delta queries.

All three operations can be performed over the entire available history of the dataset, or by specifying a time interval via a tuple in the form (START, END).

The time-agnostic-browser [2] is also released under the ISC license and can be run as a Flask application. It is organized into two macro-sections: “Explore” and “Query”. In the former, a text input accepts a URI. By submitting it, the entire history of the corresponding resource is displayed. In the latter, a text area receives a SPARQL query, which is resolved on all dataset states. Its main added value is hiding the triples and the complexity of the underlying RDF model: predicate URIs, as well as subjects and objects, appear in a human-readable format. Moreover, all the entities are displayed as links, providing shortcuts to reconstruct the history of the related resources (Figure 4).

Figure 4. Graphical user interface of an entity history reconstruction through the time-agnostic-browser.

The efficiency of time-agnostic-library was measured with two types of benchmarks [16], one on execution times and the other on the amount of computer memory (RAM) required by ten different use cases, each repeated ten times to produce significant results and avoid outliers. In light of these benchmarks, time-agnostic-library has proven effective for any materialization. Regarding structured queries, they are swift if all subjects are known or deductible. On the other hand, the presence of unknown subjects in the user’s SPARQL query involves the identification of all present and past entities that satisfy that pattern, and so requires a more significant amount of time and resources. Specifically, all materializations and the cross-version structured query with known subjects required about half a second and about 50 MB of RAM; conversely, with unknown subjects, 581 seconds and 519 MB of RAM are required. It can be concluded that the proposed software can be used effectively in all cases where the subject is known, that is, for any materialization or formulated SPARQL queries without isolated triple patterns containing unknown subjects.

Other software solutions for such problems have been proposed. Table 1 shows the list of available software to perform materializations and time-traversal queries on RDF datasets. As can be observed, time-agnostic-library is the only one to support all retrieval functionalities without requiring pre-indexing processes. This feature makes it particularly suitable for use in scenarios with large amounts of data that often change over time. Moreover, compared to the approach of Im, Lee and Kim [17] and OSTRICH [18], the OpenCitations Data Model only requires storing the current state of the dataset, rather than the original one, allowing one to query the latest version, without additional computational effort to first re-create the original version.

SoftwareVersion materializationDelta materializationSingle-version structured queryCross-version structured querySingle-delta structured queryCross-delta structured queryLive
PromptDiff [19]+++
SemVersion [20]+++
Im, Lee, & Kim, 2012 [17]+++++
R&Wbase [21]++++
x-RDF-3X [22]+++
v-RDFCSA [23]++++++
OSTRICH [18]+++
Tanon & Suchanek, 2019 [24]++++++
time-agnostic-library[1]+++++++
Table 1. Comparative between time-agnostic-library and preexisting software to achieve materializations and time traversal queries on RDF datasets. (Scroll right to see Columns 6-8).

The OpenCitations Data Model and the time-agnostic-library software are the pre-requisites that will allow OpenCitations to involve third parties, for example members of staff in academic libraries, in the submission, curation and updating of OpenCitations bibliographic and citation data. At this stage, all entities in COCI have a single snapshot — the one made at the time of creation. However, since these entities may become modified, corrected or enriched over time, it is imperative to have appropriate software tools available for use by curators. With the time-agnostic-library software and its associated time-agnostic-browser, it will be possible for a curator to explore the entire history of the changes within an RDF dataset, to know when they were made, based on which source, and by which responsible agent, thus ensuring the reliability and verifiability of data, and facilitating any necessary further changes.

References

[1] A. Massari, time-agnostic-library. 2021. Available: https://archive.softwareheritage.org/swh:1:snp:d7fd1754377f45d16afb61efc770815b5a3c8f83

[2] A. Massari, time-agnostic-browser. 2021. Available: https://archive.softwareheritage.org/swh:1:dir:337f641375cca034eda39c2380b4a7878382fc4c

[3] J. D. Fernández, A. Polleres, and J. Umbrich, ‘Towards Efficient Archiving of Dynamic Linked’, in DIACRON@ESWC, Portorož, Slovenia: Computer Science, 2015, pp. 34–49.

[4] December, ‘Provenance XG Final Report’. 2010. Available: http://www.w3.org/2005/Incubator/prov/XGR-prov-20101214/

[5] T. Käfer, A. Abdelrahman, J. Umbrich, P. O’Byrne, and A. Hogan, ‘Observing Linked Data Dynamics’, in The Semantic Web: Semantics and Big Data, vol. 7882, P. Cimiano, O. Corcho, V. Presutti, L. Hollink, and S. Rudolph, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2013, pp. 213–227. doi: 10.1007/978-3-642-38288-8_15

[6] S. L. Garfinkel, ‘Wikipedia and the Meaning of Truth’, MIT Technology Review, 2008, [Online]. Available: https://stephencodrington.com/Blogs/Hong_Kong_Blog/Entries/2009/4/11_What_is_Truth_files/Wikipedia%20and%20the%20Meaning%20of%20Truth.pdf

[7] M.-R. Koivunen and E. Miller, ‘Semantic Web Activity’, W3C, Nov. 02, 2001. https://www.w3.org/2001/12/semweb-fin/w3csw

[8] F. Orlandi and A. Passant, ‘Modelling provenance of DBpedia resources using Wikipedia contributions’, Journal of Web Semantics, vol. 9, no. 2, pp. 149–164, Jul. 2011, doi: 10.1016/j.websem.2011.03.002.

[9] P. Dooley and B. Božić, ‘Towards Linked Data for Wikidata Revisions and Twitter Trending Hashtags’, in Proceedings of the 21st International Conference on Information Integration and Web-based Applications & Services, Munich Germany, Dec. 2019, pp. 166–175. doi: 10.1145/3366030.3366048.

[10] Yago Project, ‘Download data, code, and logo of Yago projects’, Yago, 2021. https://yago-knowledge.org/downloads (accessed Sep. 24, 2021).

[11] J. Umbrich, M. Hausenblas, A. Hogan, A. Polleres, and S. Decker, ‘Towards Dataset Dynamics: Change Frequency of Linked Open Data Sources’, in Proceedings of the WWW2010 Workshop on Linked Data on the Web, Raleigh, USA, 2010. Available: http://ceur-ws.org/Vol-628/ldow2010_paper12.pdf

[12] M. Daquino, S. Peroni, and D. Shotton, ‘The OpenCitations Data Model’, p. 836876 Bytes, 2020, doi: 10.6084/M9.FIGSHARE.3443876.V7.

[13] S. Peroni, D. Shotton, and F. Vitali, ‘A Document-inspired Way for Tracking Changes of RDF Data’, in Detection, Representation and Management of Concept Drift in Linked Open Data, Bologna, 2016, pp. 26–33. Available: http://ceur-ws.org/Vol-1799/Drift-a-LOD2016_paper_4.pdf

[14] I. Heibi, S. Peroni, and D. Shotton, ‘Software review: COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations’, Scientometrics, vol. 121, no. 2, pp. 1213–1228, Nov. 2019, doi: 10.1007/s11192-019-03217-6.

[15] K. Beck, Test-driven development: by example. Boston: Addison-Wesley, 2003.

[16] A. Massari, ‘time-agnostic-library: benchmark results on execution times and RAM’. Zenodo, Oct. 05, 2021. doi: 10.5281/ZENODO.5549648.

[17] D.-H. Im, S.-W. Lee, and H.-J. Kim, ‘A Version Management Framework for RDF Triple Stores’, Int. J. Softw. Eng. Knowl. Eng., vol. 22, pp. 85–106, 2012.

[18] R. Taelman, M. V. Sande, and R. Verborgh, ‘OSTRICH: Versioned Random-Access Triple Store’, in Companion Proceedings of the Web Conference 2018, 2018, pp. 127–130. Available: https://core.ac.uk/download/pdf/157574975.pdf

[19] N. F. Noy and M. A. Musen, ‘Promptdiff: A Fixed-Point Algorithm for Comparing Ontology Versions’, in Proc. of IAAI, 2002, pp. 744–750.

[20] M. Völkel, W. Winkler, Y. Sure, S. Kruk, and M. Synak, ‘SemVersion: A Versioning System for RDF and Ontologies’, 2005.

[21] M. V. Sande, P. Colpaert, R. Verborgh, S. Coppens, E. Mannens, and R. V. Walle, ‘R&Wbase: Git for triples’, 2013.

[22] T. Neumann and G. Weikum, ‘x-RDF-3X: Fast Querying, High Update Rates, and Consistency for RDF Databases’, Proceedings of the VLDB Endowment, vol. 3, pp. 256–263, 2010.

[23] A. Cerdeira-Pena, A. Farina, J. D. Fernandez, and M. A. Martinez-Prieto, ‘Self-Indexing RDF Archives’, in 2016 Data Compression Conference (DCC), Snowbird, UT, USA, Mar. 2016, pp. 526–535. doi: 10.1109/DCC.2016.40.

[24] T. Pellissier Tanon and F. Suchanek, ‘Querying the Edit History of Wikidata’, in The Semantic Web: ESWC 2019 Satellite Events, vol. 11762, P. Hitzler, S. Kirrane, O. Hartig, V. de Boer, M.-E. Vidal, M. Maleshkova, S. Schlobach, K. Hammar, N. Lasierra, S. Stadtmüller, K. Hose, and R. Verborgh, Eds. Cham: Springer International Publishing, 2019, pp. 161–166. doi: https://doi.org/10.1007/978-3-030-32327-1_32.

Academia’s missing references

No-one is quite sure of the total number of scholarly publications within the global corpus. Indeed that number will be strongly influenced by the degree to which, in addition to books and journal articles, one includes within the definition of scholarly publications ‘grey literature’ such as reports published by official bodies, patents, etc. Consequentially, the total number of scholarly references within those publications is also unknown, and this number too will vary according to the inclusion criteria chosen. Furthermore, Crossref Event Data and similar datasets recording social media mentions of journal articles in blog posts and tweets extends the concept of a reference beyond that used in ‘conventional’ citation indexes such as COCI.

We celebrate the fact that well over one billion bibliographic citations are now openly available under CC0 waivers in NIH OCC (the National Institutes of Health Open Citation Collection) [1,2] and COCI (the OpenCitations Index of Crossref DOI-to-DOI Citations) [3]. Despite present gaps in their coverage, they include references to all the most important publications within the global corpus, because these will all have been cited multiple times.

Open references available from Crossref and other aggregators and indexes

Crossref, with over 1.6 billion open references, is the largest single source of such bibliographic metadata. Significant numbers of references are also available in a variety of other databases, repositories and indexes.

NIH OCC (the National Institutes of Health Open Citation Collection) is a merger of several citation databases, drawing on PubMed for crucial article metadata, and augmenting this with information from full-text articles that have been made freely available on the internet [1]. The CiteSeerX database, the arXiv preprint repository, and the Dryad data repository are examples of different types of infrastructure that also publish open bibliographic references, while there is further availability of article references from open aggregators such as DataCite and Wikidata. These may either use their own DOIs, DOIs from the Crossref DOI registration agency, or no DOIs at all. Either way, those references will not appear in Crossref.

What is lacking is semantic coherence and interoperability between these sources, permitting federated queries across them. This makes difficult the task of obtaining a comprehensive overview of the availability of open bibliographic references.

However, there are even more citations that are not yet freely and easily available anywhere in bulk, relating to the reference lists within publications of a number of distinct types. This blog post explores academia’s missing references – those that have not yet been documented within open freely accessible citation indexes – and what might be done to bring these into the public domain.

1 References that are closed at Crossref

Eight years ago, I wrote

“In this open-access age, it is a scandal that reference lists from journal articles — core elements of scholarly communication that permit the attribution of credit and integrate our independent research endeavours — are not readily and freely available for use by all scholars.” [4]

I stand by that statement, and, through OpenCitations [5], I have been working with colleagues to rectify the situation, as I described in my previous post. The Initiative for Open Citations (I4OC) can rightly be applauded for its part in encouraging almost all the major academic publishers who deposit references at Crossref to make them open.

The only major scholarly publisher not to be listed as an I4OC Participating Publisher is the Institute of Electrical and Electronics Engineers (IEEE), that, having deposited at Crossref reference lists for 58% of its preprints and publications, persists in keeping these deposited references closed and unavailable for indexing and re-use. It is to be hoped that IEEE will now realize that what it looses by not having references openly available outweighs any benefits it might have received from keeping them closed, and will join its fellow publishers in ensuring that its Crossref-deposited references are made open, both for its current issues and for its back-number journal articles. I thus call again upon IEEE to change its present position, as both Elsevier and the American Chemical Society had the courage to do recently, and to instruct Crossref to open all IEEE references. A single email to Crossref Support, with the instruction “Open all references”, is all it would take!

References in Crossref are either open, ‘limited’ or closed. Limited references are available to those subscribing to Crossref Metadata Plus, which includes OpenCitations, but not to the general public. Closed references are not freely available to anyone. The following table shows the number of Crossref works in each category, and the number of references within those categories.


WorksReferencesAverage references per work
Total126,627,618

Without references70,477,843
(55.7% of total)


With references56,149,775
(44.3% of total)
1,734,831,31130.9
Open49,274,155
(38.9% of total)
1,605,120,22932.6
Limited2,933,323
(2.3% of total)
66,459,70422.7
Closed3,942,297
(3.1% of total)
63,251,37816.0
Table 1. Numbers of works and their references recorded in Crossref on 31 July 2021.

2 References not deposited at Crossref for publications with Crossref DOIs

The number of works with Crossref DOIs that lack submitted references is surprisingly large. As of 31 July 2021, 70,477,843 publications (55.7% of all works recorded at Crossref) lacked deposited references (Table 1).

Crossref classifies all types of journal content, including editorials, book reviews and letters, as “journal articles”, thus some of these works without deposited references genuinely lack them. However, the majority are conventional journal articles and books with reference lists that the publishers have simply not deposited at Crossref along with the other metadata for these works.

The average number of references per Crossref work with deposited references is 30.9 (Table 1). If, to make allowance for those works that genuinely lack references, we assume a conservative average of 25 references per work for the 70,477,843 works lacking deposited references, this means that there are over 1.75 billion references within these works that have not been deposited at Crossref, and thus are not conveniently available for indexing and reuse.

These missing references relate both large publishers that have submitted references for only some of their publications, and small publishers that perhaps lack, or think they lack, the resources to deposit any reference lists in addition to the other metadata they are already sending to Crossref for each of their DOIs. However, there are several easy methods for depositing reference lists, as detailed by Crossref here. So I encourage all publishers who are supportive of Open Science to update their procedures and commence or complete the deposition of their publication reference lists at Crossref, starting with their current issues. Crossref Support will provide assistance if required. Note that a publisher does not have to subscribe to the Crossref Cited-by service to deposit its references!

3 Citations missing in COCI: open references in Crossref to publications lacking DOIs

COCI is the OpenCitations Index of Crossref DOI-to-DOI Citations, and, as the name suggests, it indexes Crossref open references from works with Crossref DOIs to other works that have DOIs [3]. It therefore does not index open references in Crossref to works that, for whatever reason, lack DOIs.

The most recent (September 2021) release of COCI, based on the August Crossref dump, contains 1,186,958,898 citations between 69,074,291 unique work, comprising 51,103,720 citing bibliographic resources bearing Crossref DOIs and 56,105,783 cited bibliographic resources. Of the cited bibliographic resources, 38,135,212 bear a DOI issued by Crossref and have open or limited references (thus also being COCI citing resources), while 17,970,571 either have a Crossref DOI but lack open or limited references or have a DOI issued by another DOI registration agency such as DataCite (thus not being COCI citing resources).

Note that in Crossref, the ratio of works with open or limited references to works without open or limited references is 0.7:1 (Table 1). However, in COCI, the ratio of cited works with Crossref DOIs containing open or limited references to all other cited works is 2.1:1. Thus works with Crossref DOIs containing open or limited references are three times more likely to be cited than works that either have a Crossref DOI but lacking open or limited references or have a DOI issued by another DOI registration agency. This is most likely because the most important journals from almost all the larger publishers now have open references. However, it is still a remarkable ratio.

Because references to works lacking DOIs are not included in COCI, the average number of bibliographic references per citing article in COCI is only 23.2, in contrast to the numbers of references to works of all types given in Table 1.

From these data, it can be seen that there are over 480 million open Crossref references to a wide variety of works lacking DOIs that OpenCitations does not index in COCI. This is because of an intentional and fundamental limitation in the structure of the Open Citation Identifier (OCI) [6], requiring both citing and cited publications to have identifiers of the same type, that lies at the heart of the functionality of our OpenCitations Indexes.

OpenCitations is currently developing a solution that, without compromising that intentional design limitation in OCIs, will nevertheless permit us to index and publish these ‘missing’ references as Linked Open Data. We will report on this development in due course.

Crossref Event Data is a Crossref service / database that records mentions of publications bearing Crossref DOIs in social media including tweets and blog posts, and in other non-traditional citation sources such as news items and Wikipedia articles. From today (Thursday 23rd September 2021), Crossref Event Data will start to include in its holdings open references from publications bearing Crossref DOIs to other publications bearing DOIs. Limited and closed references will not be included. Initially, open references from current publications will be included in Crossref Event Data, with open references from older works with DOIs being added later. In that respect it will come to resemble COCI, except that COCI also included ‘limited’ references, treats citations as first-class data entities with their own identifiers, and makes its citation data available in RDF as Linked Open Data, as well as via a REST API. Subsequently, Crossref Event Data will also record references to publications with other forms of identifier, as OpenCitations also plans to do.

4 References available on publishers’ web sites

Many publishers, particularly those of Open Access works, already make their publication reference lists openly available on their own web sites. While this is commendable, it is not sufficient, since, if scholarly references are not made available in a centralized aggregator such as Crossref from which they can be conveniently harvested in bulk for analysis and re-use, they are much more difficult to access.

Scraping references from the HTML of individual web sites is difficult, time-consuming and liable to be incomplete. While Microsoft Academic achieved considerable success in scraping references from publishers’ web sites, possible because of the special relationships these publishers have with the Microsoft search engine Bing, this service will unfortunately soon no longer be available, illustrating a problem inherent with scholarly infrastructures provided by commercial companies that do not adopt the Principles of Open Scholarly Infrastructures.

A consequence is that such publications will become increasingly ‘invisible’, as bibliographic and analytical services come to rely more and more on centrally available data.

6 References within PDFs of scholarly works lacking DOIs

There is a large but unknown quantity of reference-containing books, academic reports, patents and journal articles from publishers that, for their own good reasons, choose not to use DOIs. The text of many of these publications is already available in a marked-up machine-readable format such as JATS, used in preparation for publication, from which the reference lists could easily be extracted. Other publications are only available as PDFs, both as published Versions of Record, or as preprints deposited in a variety of preprint repositories such as arXiv or CORE. Mining reference lists out of PDFs required expertise in text mining and AI technologies, and is labour-intensive, since it usually required the tuning of extraction algorithms to handle the particular styles and formatting of individual journals, one at a time. Two stages are involved: first, the recognition and extraction of the text of the individual references from the PDF, and second the parsing of each text string into the component parts of the reference (author names, title, publication year, etc.) A considerable number of the citations within NIH-OCC have been obtained in this manner [1], commercial companies such as Lexical Intelligence specialize in this area, and publicly available software such as GROBID is available for the purpose. However, the overall task of extracting ‘missing’ academic references from the global PDF corpus is daunting in magnitude and would require a well-funded organization.

The correct way to proceed would be for each publisher to take responsibility for liberating the references of their own publications, whether or not the publications themselves are open access, and whether or not these references are already available in a marked-up machine-readable format or only within PDF documents. Then, if the publisher still chose not to use DOIs and to submit these metadata to Crossref, these references could be submitted directly to OpenCitations for aggregation and publication as Linked Open Data.

Conclusion

From the foregoing discussion it is clear that the academic community has a long way to go before the majority of scholarly citations, the products of their own labours, are openly available for analysis and re-use. We at OpenCitations are working to address these issues and to publish more of these missing citations. However, completion of the task will require a coordinated collaborative international effort.

Are you willing to be involved?

References

[1] B. Ian Hutchins et al. (2019). The NIH Open Citation Collection: A public access, broad coverage resource. PLoS Biol. 17 (10): e3000385. https://doi.org/10.1371/journal.pbio.3000385

[2] B. Ian Hutchins (2021). A tipping point for open citation data. Quantitative Science Studies 2 (2): 433–437. https://doi.org/10.1162/qss_c_00138

[3] Ivan Heibi, Silvio Peroni, David Shotton (2019). Software review: COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations. Scientometrics 121 (2): 1213-1228. https://doi.org/10.1007/s11192-019-03217-6

[4] David Shotton (2013). Open citations. Nature, 502 (7471): 295-297. http://dx.doi.org/10.1038/502295a

[5] Silvio Peroni, David Shotton (2020). OpenCitations, an infrastructure organization for open scholarship. Quantitative Science Studies, 1(1): 428-444. https://doi.org/10.1162/qss_a_00023

[6] Silvio Peroni, David Shotton (2019). Open Citation Identifier: Definition. Figshare. https://doi.org/10.6084/m9.figshare.7127816