COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations

Author(s)
Ivan Heibi – ivan.heibi2@unibo.it
Digital Humanities Advanced Research Centre (DHARC), Department of Classical Philology and Italian Studies, University of Bologna, Bologna, Italy
Silvio Peroni – silvio.peroni@unibo.it
Digital Humanities Advanced Research Centre (DHARC), Department of Classical Philology and Italian Studies, University of Bologna, Bologna, Italy
David Shotton – david.shotton@oerc.ox.ac.uk
Oxford e-Research Centre, University of Oxford, Oxford, United Kingdom

Copyright notice: This work is licensed under a Creative Commons Attribution 4.0 International License. You are free to share (i.e. copy and redistribute the material in any medium or format) and adapt (e.g. remix, transform, and build upon the material) for any purpose, even commercially, under the following terms: attribution, i.e. you must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. The licensor cannot revoke these freedoms as long as you follow the license terms.

Notes: submitted to the Resource track of the 18th International Semantic Web Conference (ISWC 2019, https://iswc2019.semanticweb.org). PDF version available on arXiv.

Abstract

In this paper, we present COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations (http://opencitations.net/index/coci). COCI is the first open citation index created by OpenCitations, in which we have applied the concept of citations as first-class data entities, and it contains more than 445 million DOI-to-DOI citation links derived from the data available in Crossref. These citations are described in RDF by means of the new extended version of the OpenCitations Data Model (OCDM). We introduce the workflow we have developed for creating these data, and also show the additional services that facilitate the access to and querying of these data by means of different access points: a SPARQL endpoint, a REST API, bulk downloads, Web interfaces, and direct access to the citations via HTTP content negotiation. Finally, we present statistics regarding the use of COCI citation data, and we introduce several projects that have already started to use COCI data for different purposes.

Introduction

The availability of open scholarly citations [21] is a public good, of significant value to the academic community and the general public. In fact, citations not only serve as an acknowledgment medium [16], but also can be characterised topologically (by defining the connected graph between citing and cited entities and its evolution over time [19]), sociologically (such as for identifying odd conduct within or elitist access paths to scientific research [18]), quantitatively by creating citation-based metrics for evaluating the impact of an idea or a person [17], and financially by defining the scholarly value of a researcher within his/her own academic community [20]. The Initiative for Open Citations (I4OC, https://i4oc.org) has dedicated the past two years to persuading publishers to provide open citation data by means of the Crossref platform (https://crossref.org), obtaining the release of the reference lists of more than 43 million articles (as of February 2019), and it is this change of behaviour by the majority of academic publishers that has permitted COCI to be created.

OpenCitations (http://opencitations.net) is a scholarly infrastructure organization dedicated to open scholarship and the publication of open bibliographic and citation data by the use of Semantic Web (Linked Data) technologies, and is a founding member of I4OC. It has created and maintains the SPAR (Semantic Publishing and Referencing) Ontologies (http://www.sparontologies.net) [22] for encoding scholarly bibliographic and citation data in RDF, and has previously developed the OpenCitations Corpus (OCC) of open downloadable bibliographic and citation data recorded in RDF [4].

In this paper, we introduce a new dataset made available a few months ago by OpenCitations, namely COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations (https://w3id.org/oc/index/coci). This dataset, launched in July 2018, is the first of the indexes proposed by OpenCitations (https://w3id.org/oc/index), in which citations are exposed as first-class data entities with accompanying properties (i.e. individuals of the class cito:Citation as defined in CiTO [7]) instead of being defined simply as relations among two bibliographic resources (via the property cito:cites). Currently COCI, contains more than 445 million DOI-to-DOI citation links made available under a Creative Commons CC0 public domain waiver, that can be accessed and queried through a SPARQL endpoint, an HTTP REST API, by means of searching/browsing Web interfaces, by bulk download in different formats (CSV and N-Triples), or by direct access via HTTP content negotiation.

The rest of the paper is organized as follows. In Section 2 we introduce some of the main RDF datasets containing scholarly bibliographic metadata and citations. In Section 3, we provide some details on the rationale and the technologies used to describe citations as first-class data entities, which are the main foundation of the development of COCI. In Section 4, we present COCI, including the workflow process developed for ingesting and exposing the open citation data available and other tools used for accessing these data. In Section 5, we show the scale of the community uptake of COCI since its launch by means of quantitative statistics on the use of its related services and by listing existing projects that are using it for specific purposes. Finally, in Section 6, we conclude the paper sketching out related and upcoming projects.

Related works

We have noticed a recent growing interest within the Semantic Web community for creating and making available RDF datasets concerning the metadata of scholarly resources, particularly bibliographic resources. In this section, we briefly introduce some of the most relevant ones.

ScholarlyData (http://www.scholarlydata.org) [1] is a project that refactors the Semantic Web Dog Food so as to keep the dataset growing in good health. It uses the Conference Ontology, an improvement version of the Semantic Web Conference Ontology, to describe metadata of documents (5,415, as of March 31, 2019), people (more than 1,100), and data about academic events (592) where such documents have been presented.

Another important source of bibliographic data in RDF is OpenAIRE (https://www.openaire.eu) [3]. Created by funding from the European Union, its RDF dataset makes available data for around 34 million research products created in the context of around 2.5 million research projects.

While important, these aforementioned datasets do not provide citation links between publications as part of their RDF data. In contrast, the following datasets do include citation data as part of the information they make available.

In 2017, Springer Nature announced SciGraph (https://scigraph.springernature.com) [2], a Linked Open Data platform aggregating data sources from Springer Nature and other key partners managing scholarly domain data. It contains data about journal articles (around 8 millions, as of March 31, 2019) and book chapters (around 4.5 millions), including their related citations, and information on around 7 million people involved in the publishing process.

The OpenCitations Corpus (OCC, https://w3id.org/oc/corpus) [4] is a collection of open bibliographic and citation data created by ourselves, harvested from the open access literature available in PubMed Central. As of March 31, 2019, it contains information about almost 14 million citation links to more than 7.5 million cited bibliographic resources.

WikiCite (https://meta.wikimedia.org/wiki/WikiCite) is a proposal, with a related series of workshops, which aims at building a bibliographic database in Wikidata [10] to serve all Wikimedia projects. Currently Wikidata hosts (as of March 29, 2019) more than 170 million citations.

Biotea (https://biotea.github.io) [5] is an RDF datasets containing information about some of the articles available in the Open Access subset of PubMed Central, that have been enhanced with specialized annotation pipelines. The last released dataset includes information extracted from 2,811 articles, including data on their citations.

Finally, Semantic Lancet [6] proposes to build a dataset of scholarly publication metadata and citations (including the specification of the citation functions) starting from articles published by Elsevier. To date it includes bibliographic metadata, abstract and citations of 291 articles published in the Journal of Web Semantics.

Indexing citations as first-class data entities

Citations are normally defined simply as links between published entities (from a citing entity to a cited entity). However, an alternative richer view is to regard each citation as a data entity in its own right, as illustrated in Figure 1. This alternative approach permits us to endow a citation with descriptive properties, such as those ones introduced in Table 1¹.

**Figure 1.** Two different ways of describing citations: as a relation between two bibliographic entities (top), or as an individual first-class data entitiy in its own right where the citing entity and the cited entity are among its attributed data.

The advantages of treating citations as first-class data entities are:

all the information regarding each citation is available in one place, since such information is defined as attributes of the citation itself;
citations become easier to describe, distinguish, count and process, and it becomes possible to distinguish separate citations within the citing entity to the cited entity, enabling one to count how many times, from which sections of the citing entity, and (in principle) for what purposes a particular cited entity is cited within the source paper;
if available in aggregate, citations described in this manner are easier to analyse using bibliometric methods, for example to determine how citation time spans vary by discipline.

We have appropriately extended the OpenCitations Data Model (OCDM, http://opencitations.net/model) [23] so as to define each citation as a first-class entity in machine-readable manner. In particular, we have used the class cito:Citation defined in the revised and expanded Citation Typing Ontology (CiTO, http://purl.org/spar/cito) [7], which is part of the SPAR Ontologies [22]. This class allows us to define a permanent conceptual directional link from the citing bibliographic entity to a cited bibliographic entity, that can be accompanied by additional ontological terms for defining specific attributes, as introduced in Table 1.

Characteristic	Description	CiTO entity
citing entity	The bibliographic entity which acts as source for the citation.	Object property `cito:hasCitingEntity`.
cited entity	The bibliographic entity which acts as target for the citation.	Object property `cito:hasCitedEntity`.
citation creation date	The date on which the citation was created. This has the same numerical value as the publication date of the citing bibliographic resource, but is a property of the citation itself. When combined with the citation time span, it permits that citation to be located in history.	Data property `cito:hasCitationCreationDate`, one of `xsd:date`, `xsd:gYearMonth`, or `xsd:gYear` as datatype value.
citation timespan	The temporal characteristic of a citation, namely the interval between the publication date of the cited entity and the publication date of the citing entity.	Data property `cito:hasCitationTimespan`, `xsd:duration` as datatype value.
type	A classification of the citation according to particular dimensions, e.g. whether or not it is a self-citation.	Property `rdf:type` associated with one or more subclasses of `cito:Citation` – in particular, for example `cito:AuthorSelfCitation` (i.e. citing and the cited entities have at least one author in common) and `cito:JournalSelfCitation` (i.e. citing and the cited entities are published in the same journal).

Table 1. List of characteristics that can be associated with a citation when it is described as first-class data entity, using the properties and classes available in CiTO for their definition in RDF.

So as to identify each citation precisely, when described as first-class data entity and included in an open dataset, we have also developed the Open Citation Identifier (OCI) [24], which is a new globally unique persistent identifier for citations. OCIs are registered in the Identifiers.org platform (https://identifiers.org/oci) and recognized as persistent identifiers for citations by the EU FREYA Project (https://www.project-freya.eu) [25]. Each OCI has a simple structure: the lower-case letters oci followed by a colon, followed by two sequences of numerals separated by a dash, where the first sequence is the identifier for the citing bibliographic resource and the second sequence is the identifier for the cited bibliographic resource. For example, oci:0301-03018 is a valid OCI for a citation defined within the OpenCitations Corpus, while oci:02001010806360107050663080702026306630509-02001010806360107050663080702026305630301 is a valid OCI for a citation included in Crossref. It is worth mentioning that OCIs are not opaque identifiers, since they explicitly encode directional relationships between identified citing and cited entities, the provenance of the citation, i.e. the database that contains it, and the type of identifiers used in that database to identify the citing and cited entities. In addition, we have created the Open Citation Identifier Resolution Service (http://opencitations.net/oci), which is a resolution service for OCIs based on the Python application oci.py available at https://github.com/opencitations/oci. Given a valid OCI as input, this resolution service is able to retrieve citation data in RDF (either as RDF/XML, Turtle or JSON-LD), or in Scholix, JSON or CSV formats. A more detailed explanation of OCIs and related material is available in [24].

At OpenCitations, we define an open citation index as a dataset containing citations that complies with the following requirements:

the citations contained are all open, according to the definition provided in [21];
the citations are all treated as first-class data entities;
each citation is identified by an Open Citation Identifier (OCI) [24];
the citation data are recorded in RDF according to the OpenCitations Data Model (OCDM) [23], where the OCI of a citation is embedded in the IRI defining it in RDF;
each citation defines the attributes shown in Table 1.

COCI: ingestion workflow, data, and services

COCI, the OpenCitations Index of Crossref open DOI-to-DOI references, is the first citation index to be published by OpenCitations, in which we have applied the concept of citations as first-class data entities, introduced in the previous section, to index the contents of one of the major open databases of scholarly citation information, namely Crossref (https://crossref.org), and to render and make available this information in machine-readable RDF under a CC0 waiver. Crossref contains metadata about publications (mainly academic journal articles) that are identified using Digital Object Identifiers (DOIs). Out of more than 100 million publications recorded in Crossref, Crossref also stores the reference lists of more than 43 million publications deposited by the publishers. Many of these references are to other publications bearing DOIs that are also described in Crossref, while others are to publications that lack DOIs and do not have Crossref descriptions. Crossref organises such publications with associated reference lists according to three categories: closed, limited and open. These categories to publications for which the reference lists are not visible to anyone outside the Crossref Cited-by membership, are visible only to them and to Crossref Metadata Plus members, or are visible to all users, respectively².

**Figure 2.** The diagram of the data model adopted to define the new class for defining citations as first-class data entities, which forms part of the OpenCitations Data Model. This model uses terms from the Citation Typing Ontology (CiTO, http://purl.org/spar/cito) for describing the data, and from the Provenance Ontology (PROV-O, http://www.w3.org/ns/prov) to define the citation’s provenance.

Followed the first release of COCI on June 4, 2018, the most recent version of COCI, released on November 12, 2018, contains more that 445 million DOI-to-DOI citations included in the open and the limited datasets of Crossref reference data³. All the citation data in COCI and their provenance information, described according the Graffoo diagram [27] presented in Figure 2, are included in two distinct graphs – https://w3id.org/oc/index/coci/ and https://w3id.org/oc/index/coci/prov/ respectively – released under a CC0 waiver, and compliant with the FAIR data principles [26].

An example of a citation included in COCI is shown in the following excerpt (in Turtle), where the OCI is embedded as part of the IRI of the citation (without the oci: prefix) after the ci/ (meaning citation according to the OpenCitations Data Model [23]):

@prefix cito: <http://purl.org/spar/cito/> .
@prefix prov: <http://www.w3.org/ns/prov#> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .

<https://w3id.org/oc/index/coci/ci/02001010806360107050663080702026306630509-02001010806360107050663080702026305630301>
  a 
    cito:Citation,
    cito:JournalSelfCitation ;
  cito:hasCitationCreationDate "2013"^^xsd:gYear ;
  cito:hasCitationTimeSpan "P1Y"^^xsd:duration ;
  cito:hasCitingEntity <http://dx.doi.org/10.1186/1756-8722-6-59> ;
  cito:hasCitedEntity <http://dx.doi.org/10.1186/1756-8722-5-31> ;
  prov:generatedAtTime "2018-11-01T05:47:54+00:00"^^xsd:dateTime ;
  prov:hadPrimarySource <https://api.crossref.org/works/10.1186/1756-8722-6-59> ;
  prov:wasAttributedTo <https://w3id.org/oc/index/coci/prov/pa/1> .

In the following subsections we introduce the ingestion workflow developed for creating COCI, we provides some figures on the citations it contains, and we list the resources and services we have made available to permit access to and querying of the dataset.

Ingestion workflow

We processed all the data included in the October 2018 JSON dump of Crossref data, available to all the Crossref Metadata Plus members. The ingestion workflow, summarised in Figure 3, was organised in four distinct phases, and all the related scripts developed and used are released as open source code according to the ISC License and downloadable from the official GitHub repository of COCI at https://github.com/opencitations/coci.

**Figure 3.** A flowchart scheme describing the workflow to build COCI. It is divided in four phases: (1) global data generation, (2) CSV generation, (3) conversion into RDF, and (4) updating the triplestore.

Phase 1: global data generation. We parse and process the entire Crossref bibliographic database to extract all the publications having a DOI and their available list of references. Through this process three datasets are generated, which are used in the next phase:

Dates, the publication dates of all the bibliographic entities in Crossref and of all their references if they explicitly specify a DOI and a publication date as structured data – e.g. see the fields DOI and year in the array reference in https://api.crossref.org/works/10.1007/978-3-030-00668-6_8. Where the same DOI is encountered multiple times, e.g. as a proper item indexed in Crossref and also as a reference in the reference list of another article deposited in the Crossref, we use the full publication date defined in the indexed item.
ISSN: the ISSN (if any) and publication type (journal-article, book-chapter, etc.) of each bibliographic entity identified by a DOI indexed in Crossref.
ORCID: the ORCIDs (if any) associated with the authors of each bibliographic entity identified by a DOI indexed in Crossref.

Phase 2: CSV generation. We generate a CSV file such that each row represents a particular citation between a citing entity and a cited entity according to the data available in the Crossref dump, by looking at the DOI identifying the citing entity and all the DOIs specified in the reference list of such a citing entity according to the Crossref data. In particular, we execute the following four steps for each citation identified:

We generate the OCI for the citation by encoding the DOIs of the citing and cited entities into numerical sequences using the lookup table available at https://github.com/opencitations/oci/blob/master/lookup.csv, which are prefixed by the supplier prefix 020 to indicate Crossref as the source of the citation.
We retrieve the publication date of the citing entity from the Dates dataset and assign it as citation creation date.
We retrieve the publication date of the cited entity (from the Dates dataset) and we use it, together with the publication date of the citing entity retrieved in the previous step, to calculate the citation timespan.
We use the data contained in the ISSN and ORCID datasets to establish whether the citing and cited entity have been published in the same journal and/or have at least one author in common, and in these cases we assign the appropriate self-citation type(s) to the citation.

Simultaneously with the creation of the CSV file of citation data, we generate a second CSV file containing the provenance information for each citation (identified by its OCI generated in the aforementioned Step 1). These provenance data include the agent responsible for the generation of the citation, the Crossref API call that refers to the data of the citing bibliographic entity containing the reference used to create the citation, and the creation date of the citation.

Phase 3: converting into RDF. The CSV files generated in the previous phase are then converted into RDF according to the N-Triples format, following the OWL model introduced in Figure 2, where the DOIs of the citing and cited entities become DOI URLs starting with http://dx.doi.org/⁴, while the IRI of the citation includes its OCI (without the oci: prefix), as illustrated in the example given in the previous section.

Phase 4: updating the triplestore. The final RDF files generated in Phase 3 are used to update the triplestore used for the OpenCitations Indexes.

Data

COCI was first created and released on July 4, 2018, and most recently updated on November 12, 2018. Currently, it contains 445,826,118 citations between 46,534,705 bibliographic entities. These are stored by means of 2,259,134,894 RDF statements (around 5 RDF statements per citation) for describing the citation data, and 1,337,478,354 RDF statements (3 statements per citation) for describing the related provenance information. Of the citations stored, 29,755,045 (6.7%) are journal self-citations, while 250,991 (0.06%) are author self-citations. The number of identified author self-citations, based on author ORCIDs, is a significant underestimate of the true number, mainly due to the sparsity of the data concerning the ORCID author identifiers within the Crossref dump. Journal entities (i.e. journals, volumes, issues, and articles) are the type of the bibliographic entities that are mostly cited, with over 420 million citations.

We also classify the cited documents according to their publishers – Table 2 shows the ten top publishers of citing and cited documents, calculated by looking at the DOI prefixes of the entities involved in each citation. As we can see, Elsevier is by far the publisher having the majority of cited documents. It is also the largest publisher that is not participating in the Initiative for Open Citations by making its publications’ reference lists open at Crossref – which is highlighted by the very limited amount of outgoing citations recorded in COCI. Its present refusal to open its article reference lists in Crossref, contrary to the practice of most of the major scholarly publishers, is contributing significantly to the invisibility of Elsevier’s own publications within the corpora of open citation data such as COCI that are increasingly being used by the scholarly community for discovery, citation network visualization and bibliometric analysis, as we introduce below in the section entitled Section 5.

Publisher	Outgoing citations	Incoming citations
Springer Nature	79,860,827	52,257,862
Wiley	76,819,685	48,174,542
Elsevier	2,853,739	96,310,027
Informa UK Limited	41,433,917	14,975,989
Institute of Electrical and Electronics Engineers (IEEE)	30,114,985	20,940,703
American Physical Society (APS)	15,729,297	16,065,862
SAGE Publications	15,933,805	7,915,082
Ovid Technologies (Wolters Kluwer Health)	9,971,274	12,840,293
Oxford University Press (OUP)	9,891,000	11,466,659
AIP Publishing	10,130,022	8,455,097

Table 2. A classification of the COCI citations according to the publishers of the cited (incoming citations) and citing (outgoing citations) documents. The table shows the top ten publishers by the overall amount of incoming and outgoing to/from their published works. Those publishers shown in italics are not participating in the Initiative for Open Citations by making their publications’ reference lists open at Crossref – see https://i4oc.org for additional information.

Resources and services

The citation data in COCI can be accessed in a variety of convenient ways, listed as follows.

Open Citation Index SPARQL endpoint. We have made available a SPARQL endpoint for all the indexes released by OpenCitations, including COCI, which is available at https://w3id.org/oc/index/sparql. When accessed with a browser, it shows a SPARQL endpoint editor GUI generated with YASGUI [8]. Of course, this SPARQL endpoint can additionally be queried using the REST HTTP protocol, e.g. via curl. In order to access to COCI data, the graph https://w3id.org/oc/index/coci/ must be specified in the SPARQL query.

COCI REST API. Citation data in COCI can be retrieved by using the COCI REST API, available at https://w3id.org/oc/index/coci/api/v1. The rationale of making a REST API available in addition to the SPARQL endpoint was to provide convenient access to the the citation data included in COCI for Web developers and users who are not necessarily experts in Semantic Web technologies. This REST API, as are all the other REST APIs made available by OpenCitations, has been implemented by means of RAMOSE, the Restful API Manager Over SPARQL Endpoints (https://github.com/opencitations/ramose), which is a Python application that allows one to simply create a REST API over any SPARQL endpoint by means of a simple configuration file that execute a SPARQL query dependently of the particular API call specified. The configuration file for the COCI API is available at https://github.com/opencitations/api/blob/master/coci_v1.hf. Currently, the COCI REST API makes available four operations, that will retrieve either (a) the citation data for all the references of a given DOI (operation: references), or (b) the citation data for all the citations received by a given DOI (operation: citations), or (c) the citation data for the citation identified by an OCI (operation: citation), or (d) the metadata for the articles identified by the specified DOIs (operation: metadata). It is worth mentioning that the latter operation strictly depends on live API calls to external services, namely the Crossref API (https://api.crossref.org), the DataCite API (https://api.datacite.org), and the Unpaywall API (http://api.unpaywall.org), to gather the metadata of the requested articles, such as the title, the authors, and the journal name, that are not explicitly included within the OpenCitations Index triplestore.

Searching and browsing interfaces. We have additionally developed a user-friendly text search interface (https://w3id.org/oc/index/search), and a browsing interface (e.g. https://w3id.org/oc/index/browser/coci/ci/02001010806360107050663080702026306630509-02001010806360107050663080702026305630301), that can be used to search citation data in all the OpenCitations Indexes, including COCI, and to visualise and browse them, respectively. These two interfaces have been developed by means of OSCAR, the OpenCitations RDF Search Application (https://github.com/opencitations/oscar) [9], and LUCINDA, the OpenCitations RDF Resource Browser (https://github.com/opencitations/lucinda), that provide a configurable layer over SPARQL endpoints that permit one easily to create Web interfaces for querying and visualising the results of SPARQL queries.

Data dumps. All the citation data and provenance information in COCI are available as dumps stored in Figshare (https://figshare.com) in both CSV and N-Triples formats, while a dump of the whole triplestore is available on The Internet Archive (https://archive.org). The links to these dumps are available on the download page of the OpenCitations website (http://opencitations.net/download#coci).

Direct HTTP access. All the citation data in COCI can be accessed directly by means of the HTTP IRIs of the stored resources (via content negotiation, e.g. https://w3id.org/oc/index/coci/ci/02001010806360107050663080702026306630509-02001010806360107050663080702026305630301).

Quantifying the use of COCI citation data

In the past months, we have monitored the accesses to COCI data since its launch in July 2018. The statistics and graphics we show in this section highlight two different aspects: the quantification of the use of COCI data – and related services – and the community uptake, i.e. the use of COCI data for specific reuses within cross-community projects and studies. All the data of the charts described in this section are freely available for download from Figshare [15].

Quantitative analysis

Figure 4 shows the number of accesses made between July 2018 and February 2019 (inclusive) to the various COCI services described above – the search/browse interfaces, the REST API, SPARQL queries, and others (e.g. direct HTTP access to particular citations and visits to COCI webpages in the OpenCitations website). We have excluded from all these counts all accesses made by automated agents and bots. As shown, the REST API is, by far, the most used service, with extensive usage recorded in the last four months, following the announcement of the second release of COCI. This is reasonable, considering that the REST API has been developed exactly for accommodating the needs of generic Web users and developers, including (and in particular) those who are not expert in Semantic Web technologies. There is just one exception in November 2018, where the SPARQL endpoint was used to retrieve quite a large amount of citation data. After further investigation, we noticed a large proportion of the SPARQL calls were coming from a single source (according to the IP data stored in our log), which probably collected citation data for a specific set of entities.

**Figure 4.** The number of accesses to COCI-related services since July 2018 to February 2019. The scale used in the y-axis is logarithmic.

Figure 5 shows a particular cut of the figures given in Figure 4, which focuses on the REST API accesses only. In particular, we analysed which operations of the API were used the most. According to these figures, the most used operation is metadata (which was first introduced in the API in August 2018) which allows one to retrieve all the metadata describing certain publications. In contrast to the other API operations, this metadata search accepts one or more DOIs as input. The least used operation was citation, which allows one to retrieve citation data given an OCI, which should not be surprising, considering the currently limited knowledge of this new identifier system for citations.

**Figure 5.** The number of access made to each different COCI REST-API operation since the release of COCI on July 2018. Classified into 4 categories (requested resource): references, citations, citation, and metadata, as defined in the text.Note again the logarithmic scale of the y-axis.

In addition, we have also retrieved data about the views and downloads (as of March 29, 2019) of all the dumps uploaded to Figshare and to the Internet Archive. The CSV data dump received 1,321 views and 454 downloads, followed by the N-Triples data dump with 316 views and 93 downloads. The CSV provenance information dump has 166 and 127 downloads, while the N-Triples provenance information dump had 95 views and 34 downloads. Finally, the least accessed dump was that of the entire triplestore available in the Internet Archive, uploaded for the very first time in November 2018, that had only 3 views.

Community uptake

The data in COCI has been already used in various projects and initiatives. In this section, we list all the tools and studies doing this of which we are aware.

VOSviewer (http://www.vosviewer.com) [11] is a software tool, developed at the Leiden University’s Centre for Science and Technology Studies (CWTS), for constructing and visualizing bibliometric networks, which may include journals, researchers, or individual publications, and may be constructed based on citation, bibliographic coupling, co-citation, and co-authorship relations. Starting from version 1.6.10 (released on January 10, 2019), VOSviewer can now directly use citation data stored in COCI, retrieved by means of the COCI REST API.

Citation Gecko (http://citationgecko.com) is a novel literature mapping tool that allows one to map a research citation network using some initial seed articles. Citation Gecko is able to leverage citation links between seed papers and other papers to highlight papers of possible interest to the user, for which it uses COCI data (accessed via the REST API) to generate the citation network.

OCI Graphe (https://dossier-ng.univ-st-etienne.fr/scd/www/oci/OCI_graphe_accueil.html) is a Web tool that allows one to search articles by means of the COCI REST API, that are then visualised in a graph showing citations to the retrieved articles. It enriches this visualisation by adding additional information about the publication venues, publication dates, and other related metadata.

Zotero [12] is a free, easy-to-use tool to help users collect, organize, cite, and share research. Recently, the Open Citations Plugin for Zotero (https://github.com/zuphilip/zotero-open-citations) has been released, which allows users to retrieve open citation data extracted from COCI (via its REST API) for one or more articles included in a Zotero library.

COCI data, downloaded from the CSV dump available on Figshare, have been also used in at least two bibliometric studies. In particular, during the LIS Bibliometrics 2019 Event, Stephen Pearson presented a study (https://blog.research-plus.library.manchester.ac.uk/2019/03/04/using-open-citation-data-to-identify-new-research-opportunities/) run on publications by scholars at the University of Manchester which used COCI to retrieve citations between these publications so as to investigate possible cross-discipline and cross-department potential collaborations. Similarly, COCI data were used to conduct an experiment on the latest Italian Scientific Habilitation [13] (the national exercise that evaluates whether a scholar is appropriate to receive an Associate/Full Professorship position in an Italian university), which aimed at trying to replicate part of the outcomes of this evaluation exercise for the Computer Science research field by using only open scholarly data, including the citations available in COCI, rather than citation data from subscription services.

Conclusions

In this paper, we have introduced COCI, the OpenCitations Index of Crossref open DOI-to-DOI citations. After an initial introduction of the notion of citations as first-class data entities, we have presented the ingestion workflow that has been implemented to create COCI, have detailed the data COCI contains, and have described the various services and resources that we have made available to access COCI data. Finally, we have presented some statistics about the use of COCI data, and have mentioned the tools and studies that have adopted COCI in recent months.

COCI is just the first open citations index that OpenCitations will make available. Using the experience we have gathered by creating it, we now plan the release of additional indexes, so as to extend the coverage of open citations available through the OpenCitations infrastructure. The first of these, recently released, is CROCI (https://w3id.org/oc/index/croci) [14], the Crowdsourced Open Citations Index, which contains citations deposited by individuals. CROCI is designed to permit scholars proactively to fill the open citations gap in COCI resulting from four causes: (a) the failure of many publishers using Crossref DOIs to deposit reference lists of their publications at Crossref, (b) the failure of some publishers that do deposit their reference lists to make these reference lists open, in accordance with the recommendations of the Initiative for Open Citations; (c) the absence from ~11% of Crossref reference metadata of the DOIs for cited articles which in fact have been assigned DOIs (https://www.crossref.org/blog/underreporting-of-matched-references-in-crossref-metadata/), a problem that Crossref are currently working hard to rectify; and (d) the existence of citations to published entities that lack Crossref DOIs. In the near future, we plan to extend the number of indexes by harvesting citations from other open datasets including Wikidata (https://www.wikidata.org), DataCite (https://datacite.org), and Dryad (https://datadryad.org). In addition, we plan to extend and generalise the current software developed for COCI, so as to facilitate most frequent updates of the indexes.

Acknowledgements

We gratefully acknowledge the financial support provided to us by the Alfred P. Sloan Foundation for the OpenCitations Enhancement Project (grant number G‐2017‐9800).

References

Nuzzolese, A. G., Gentile, A. L., Presutti, V., Gangemi, A. (2016). Conference Linked Data: The ScholarlyData project. In Proceedings of the 15^th International Semantic Web Conference (ISWC 2015): 150-158. DOI: https://doi.org/10.1007/978-3-319-46547-0_16
Hammond, T., Pasin, M., & Theodoridis, E. (2017). Data integration and disintegration: Managing Springer Nature SciGraph with SHACL and OWL. In International Semantic Web Conference (Posters, Demos & Industry Tracks). http://ceur-ws.org/Vol-1963/paper493.pdf
Alexiou, G., Vahdati, S., Lange, C., Papastefanatos, G., Lohmann, S. (2016). OpenAIRE LOD services: scholarly communication data as linked data. In Semantics, Analytics, Visualization. Enhancing Scholarly Data: 45-50. DOI: https://doi.org/10.1007/978-3-319-53637-8_6
Peroni, S., Shotton, D., Vitali, F. (2017). One year of the OpenCitations Corpus – releasing RDF-based scholarly citation data into the public domain. In Proceedings of the 16^th International Semantic Web Conference (ISWC 2017): 184-192. DOI: https://doi.org/10.1007/978-3-319-68204-4_19
Garcia, A., Lopez, F., Garcia, L., Giraldo, O., Bucheli, V., Dumontier, M. (2018). Biotea: semantics for Pubmed Central. PeerJ, 6: e4201. DOI: https://doi.org/10.7717/peerj.4201
Bagnacani, A., Ciancarini, P., Di Iorio, A., Nuzzolese, A. G., Peroni, S., Vitali, F. (2014). The Semantic Lancet Project: A Linked Open Dataset for Scholarly Publishing. In EKAW 2014 Satellite Events: 101-105. DOI: https://doi.org/10.1007/978-3-319-17966-7_10
Silvio Peroni, David Shotton (2012). FaBiO and CiTO: ontologies for describing bibliographic resources and citations. Web Semantics, 17: 33-34. DOI: https://doi.org/10.1016/j.websem.2012.08.001
Rietveld, L., Hoekstra, R. (2017). The YASGUI family of SPARQL clients Semantic Web, 8(3): 373-383. DOI: https://doi.org/10.3233/SW-150197
Heibi, I., Peroni, S., Shotton, D. (2018). OSCAR: A Customisable Tool for Free-Text Search over SPARQL Endpoints. In Semantics, Analytics, Visualization: 121-137. DOI: https://doi.org/10.1007/978-3-030-01379-0_9
Erxleben, F., Günther, M., Krötzsch, M., Mendez, J., Vrandečić, D. (2014). Introducing Wikidata to the linked data web. In Proceedings of the 13^th International Semantic Web Conference (ISWC 2013): 50-65. DOI: https://doi.org/10.1007/978-3-319-11964-9_4
van Eck, N., & Waltman, L. (2009). Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics, 84(2), 523-538. DOI: https://doi.org/10.1007/s11192-009-0146-3
Ahmed, K. M., Al Dhubaib, B. (2011). Zotero: A bibliographic assistant to researcher. Journal of Pharmacology and Pharmacotherapeutics, 2(4), 303. DOI: https://doi.org/
10.4103/0976-500X.85940
Di Iorio, A., Peroni, S., Poggi, F. (2019). Open data to evaluate academic researchers: an experiment with the Italian Scientific Habilitation. (To appear) Proceedings of the 17th International Conference on Scientometrics and Informetrics (ISSI 2019). https://arxiv.org/abs/1902.03287
Heibi, I., Peroni, S., Shotton, D. (2019). Crowdsourcing open citations with CROCI – An analysis of the current status of open citations, and a proposal. (To appear) Proceedings of the 17th International Conference on Scientometrics and Informetrics (ISSI 2019). https://arxiv.org/abs/1902.02534
Heibi, I., Peroni, S., Shotton, D. (2019). Usage statistics of COCI data. Figshare. DOI: https://doi.org/10.6084/m9.figshare.7873559
Newton, I. (1675). Isaac Newton letter to Robert Hooke – Cambridge, 5 February 1675. https://digitallibrary.hsp.org/index.php/Detail/objects/9792 (last visited 23 March 2019)
Schiermeier, Q. (2017). Initiative aims to break science’s citation paywall. Nature. DOI: https://doi.org/10.1038/nature.2017.21800
Sugimoto, C. R., Waltman, L., Larivière, V., van Eck, N. J, Boyack, K. W., Wouters, P., de Rijcke, S. (2017). Open citations: A letter from the scientometric community to scholarly publishers. ISSI Society. http://issi-society.org/open-citations-letter (last visited 23 March 2019)
Chawla, D. S. (2017). Now free: citation data from 14 million papers, and more might come. Science. https://www.sciencemag.org/news/2017/04/now-free-citation-data-14-million-papers-and-more-might-come (last visited 23 March 2019)
Molteni, M. (2017). Tearing Down Science’s Citation Paywall, One Link at a Time. Wired. https://www.wired.com/2017/04/tearing-sciences-citation-paywall-one-link-time/ (last visited 23 March 2019)
Peroni, S., Shotton, D.. (2018). Open Citation: Definition. Figshare. DOI: https://doi.org/10.6084/m9.figshare.6683855
Peroni, S., Shotton, D. (2018). The SPAR Ontologies. In Proceedings of the 17th International Semantic Web Conference (ISWC 2018): 119-136. DOI: https://doi.org/10.1007/978-3-030-00668-6_8
Peroni, S., Shotton, D. (2018). The OpenCitations Data Model. Figshare. DOI: https://doi.org/10.6084/m9.figshare.3443876
Peroni, S., Shotton, D. (2019). Open Citation Identifier: Definition. Figshare. DOI: https://doi.org/10.6084/m9.figshare.7127816
Ferguson, C., McEntrye, J., Bunakov, V., Lambert, S., van der Sandt, S., Kotarski, R., … McCafferty, S. (2018). Survey of Current PID Services Landscape (Deliverable No. D3.1). Retrieved from FREYA project (EC Grant Agreement No 777523) website: https://www.project-freya.eu/en/deliverables/freya_d3-1.pdf
Wilkinson, M. D., Dumontier, M., Aalbersberg, Ij. J., Appleton, G., Axton, M., Baak, A., … Mons, B. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3, 160018. DOI: https://doi.org/10.1038/sdata.2016.18
Falco, R., Gangemi, A., Peroni, S., Shotton, D., Vitali, F. (2014). Modelling OWL Ontologies with Graffoo. In The Semantic Web: ESWC 2014 Satellite Events: 320–325. DOI: https://doi.org/10.1007/978-3-319-11955-7_42

Footnotes

1. An in-depth description about the definition and use of citations as first-class data entities can be found at https://opencitations.hypotheses.org/816.

2. Additional information on this classification of Crossref reference lists is available at https://www.crossref.org/reference-distribution/.

3. We have access to the limited dataset since we are members of the Crossref Metadata Plus plan.

4. We are aware that the current practice for DOI URLs is to use the base https://doi.org/ instead of http://dx.doi.org/. However, when one tries to resolve a DOI URL owned by Crossref by specifying an RDF format (e.g. Turtle) in the accept header of the request, the bibliographic entity is actually defined using the old URL structure starting with http://dx.doi.org/. For this reason, since COCI is derived entirely from Crossref data, we decided to stay with the approach currently used by Crossref.

Crowdsourcing open citations with CROCI

Crowdsourcing open citations with CROCI
An analysis of the current status of open citations, and a proposal

Keywords: Open citations; COCI; CROCI; Crossref; I4OC; OpenCitations

Notes: submitted as research-in-progress paper to the 17th International Conference on Scientomentrics and Informetrics (ISSI 2019, https://www.issi2019.org). PDF version available on arXiv.

Abstract

In this paper, we analyse the current availability of open citations data in one particular dataset, namely COCI (the OpenCitations Index of Crossref open DOI-to-DOI citations; http://opencitations.net/index/coci) provided by OpenCitations. The results of these analyses show a persistent gap in the coverage of the currently available open citation data. In order to address this specific issue, we propose a strategy whereby the community (e.g. scholars and publishers) can directly involve themselves in crowdsourcing open citations, by uploading their citation data via the OpenCitations infrastructure into our new index, CROCI, the Crowdsourced Open Citations Index.

Introduction

The availability of open scholarly citations is a public good, which is of intrinsic value to the academic world as a whole (Shotton, 2013; Peroni et al. 2015; Shotton, 2018), and is particularly crucial for the scientometrics and informetrics community, since it supports reproducibility (Sugimoto et al., 2017) and enables fairness in research by removing such citation data from behind commercial paywalls (Schiermeier, 2017). Despite the positive early outcome of the Initiative for Open Citations (I4OC, https://i4oc.org), namely that almost all major scholarly publishers now release their publication reference lists, with the result that more than 500 million citations are now open via the Crossref API (https://api.crossref.org), and despite the related ongoing efforts of sister infrastructures and initiatives such as OpenCitations (http://opencitations.net) and WikiCite/Wikidata (https://www.wikidata.org), many scholarly citations are not freely available. While these initiatives have the potential to disrupt the traditional landscape of citation availability, which for the past half-century has been dominated by commercial interests, the present incomplete coverage of open citation data is one of the most significant impediments to open scholarship (van Eck et al., 2018).

In this work, we analyse the current availability of open citations data (Peroni and Shotton, 2018) within one particular dataset, namely COCI (the OpenCitations Index of Crossref open DOI-to-DOI citations; http://opencitations.net/index/coci). This dataset is provided by OpenCitations, a scholarly infrastructure organization dedicated to open scholarship and the publication of open bibliographic and citation data by the use of Semantic Web (Linked Data) technologies. Launched in July 2018, COCI is the first of the Indexes proposed by OpenCitations (http://opencitations.net/index) in which citations are exposed as first-class data entities with accompanying properties. It has already seen widespread usage (over nine hundred thousands API calls since launch, with half of these in January 2019), and has been adopted by external services such as VOSviewer (van Eck and Waltman, 2010).

In particular, in this paper we address the following research questions (RQs):

What is the ratio between open citations vs. closed citations within each category of scholarly entities included in COCI (i.e. journals, books, proceedings, datasets, and others)?
Which are the top twenty publishers in terms of the number of open citations received by their own publications, according to the citation data available in COCI?
To what degree are the publishers highlighted in the previous analysis themselves contributing to the open citations movement, according to the data available in Crossref?

The results of these analyses show a persistent gap in the coverage of the currently available open citation data. To address this specific issue, we have developed a novel strategy whereby members of the community of scholars, authors, editors and publishers can directly involve themselves in crowdsourcing open citations, by uploading their citation data via the OpenCitations infrastructure into our new index, CROCI, the Crowdsourced Open Citations Index.

Methods and material

To answer the RQs mentioned above, we used open data and technologies coming from various parties. Specifically, the open CC0 citation data we used came from the CSV dump of most recent release of COCI dated 12 November 2018 (OpenCitations, 2018), which contains 449,840,503 DOI-to-DOI citation links between 46,534,705 distinct bibliographic entities. The Crossref dump we used for the production of this most recent version of COCI was dated 3 October 2018, and included all the Crossref citation data available at that time in both the ‘open’ dataset (accessible by all) and the ‘limited’ dataset (accessible only to users of the Crossref Cited-by service and to Metadata Plus members of Crossref, of which OpenCitations is one – for details, see https://www.crossref.org/reference-distribution/).

We additionally extracted information about the number of closed citations to each of the 99,444,883 DOI-identified entities available in the October Crossref dump. This number was calculated by subtracting the number of open citations to each entity available within COCI from the value “is-referenced-by-count” available in the Crossref metadata for that particular cited entity, which reports all the DOI-to-DOI citation links that point to the cited entity from within the whole Crossref database (including those present in the Crossref ‘closed’ dataset).

Furthermore, we extracted the particular publication type of each entity, so as to identify it either as a journal article, or as a book chapter, etc. We determined these publication types for all the DOI-identified entities available in the Crossref dump we used. We then identified the publisher of each entity, by querying the Crossref API using the entity’s DOI prefix. This allowed us to group the number of open citations and closed citations to the articles published by that particular publisher, and to determine the top twenty publishers in terms of the number of open citations that their own publications had received.

Finally, we again queried the Crossref API, this time using the DOI prefixes of the citing entities, to check the participation of these top twenty publishers in terms of the number of open citations they were themselves publishing in response to the open citation movement sponsored by I4OC. Details of all these analyses are available online in CC0 (Heibi et al., 2019).

Results

First (RQ1) we determined the numbers of open citations and closed citations received by the entities in the Crossref dump. All the entity types retrieved from Crossref were aligned to one of following five categories: journal, book, proceedings, dataset, other, as illustrated in (Heibi et al., 2019). The outcomes are summarised in Figure 1, where it is evident that the number of open citations available in COCI is always greater than the number of closed citations to these entities to which COCI does not have access, for each of the publication categories considered, with the categories proceedings and dataset having the largest ratios.

Figure 1. The number of open citations (available in COCI) vs. closed citations (according to Crossref data) of the cited entities within COCI, analyzed and grouped according to five distinct categories. [Note that the vertical axis has a logarithmic scale].

Analysis of the Crossref data show that there are in total ~4.1 million DOIs that have received no open citations and at least one closed citation. Conversely, there are ~10.7 million DOIs that have received no closed citations and at least one open citation in COCI. Most of the papers in both these categories have received very few citations.

The outcome of the second analysis (RQ2) shows which publishers are receiving the most open citations. To this end, we considered all the open citations recorded in COCI, and compared them with the number of closed citations to these same entities recorded in Crossref. Figure 2 shows the top twenty publishers that received the greatest number of open citations. Elsevier is the first publisher according to this ranking, but it also records the highest number of closed citations received (~97M vs. ~105.5M). The highest ratio in terms of open citations vs. closed citations was recorded by IEEE publications (ratio 6.25 to 1), while the lowest ratio was for the American Chemical Society (ratio 0.73 to 1).

Figure 2. The top twenty publishers sorted in decreasing order according to the number of open citations the entities they published have received, according to the open citation data within COCI. We accompany this count with the number of closed citations to the entities published by each of them according to the values available in Crossref, the total numbers of citations to these publishers’ entities, and the percentages of these totals that are open or closed.

Considering the twenty publishers listed in Figure 2, we wanted additionally to know their current support for the open citation movement (RQ3). The results of this analysis (made by querying the Crossref API on 24 January 2019) are shown in Figure 3. Among the top ten publishers shown in Figure 2, i.e. those who themselves received the largest numbers of open citations, only five, namely Springer Nature, Wiley, the American Physical Society, Informa UK Limited, and Oxford University Press, are participating actively in the open publication of their own citations through Crossref.

Figure 3. The contributions to open citations made by the twenty publishers listed in Figure 2, as of 24 January 2018, according to the data available through the Crossref API. The counts listed in this table refers to the number of publications for which each publisher has submitted metadata to Crossref that include the publication’s reference list. The categories closed, limited and open refer to publications for which the reference lists are not visible to anyone outside the Crossref Cited-by membership, are visible only to them and to Crossref Metadata Plus members, or are visible to all, respectively. Additional information on this classification of Crossref reference lists is available at https://www.crossref.org/reference-distribution/. The final column in the table shows the total number of publications for which the publisher has submitted metadata to Crossref, whether or not those metadata include the reference lists of those publications.

It is noteworthy that JSTOR contributes very few references to Crossref, while the many citations directed towards its own holdings place JSTOR twelfth in the list of publishers receiving open citations (Figure 2). However, as the last column of Figure 3 shows, all the major publishers listed here are failing to submit reference lists to Crossref for a large number of the publications for which they submit metadata, that number being the difference between the value in the last column for that publisher and the combined values in the preceding three columns. JSTOR is the worst in this regard, submitting references with only 0.53% of its deposits to Crossref, while the American Physical Society is the best, submitting references with 96.54% of its publications recorded in Crossref.

Additional information about these analyses, including the code and the data we have used to compute all the figures, is available as a Jupyter notebook at https://github.com/sosgang/pushing-open-citations-issi2019/blob/master/script/croci_nb.ipynb.

It should be stressed that a very large number of potentially open citations are totally missing in the Crossref database, and consequently from COCI, for the simple reason that many publishers, particularly smaller ones with limited technical and financial resources, but also all the large ones shown in Figure 3 and most of the others, are simply not depositing with Crossref the reference lists for any or all of their publications.

Discussion

According to the data retrieved, the open DOI-to-DOI citations available in COCI exceed the number of closed DOI-to-DOI citations recorded in Crossref for every publication category, as shown in Figure 1. The journal category is the one receiving the most open citations overall, as expected considering the historical and present importance of journals in most areas of the scholarly ecosystem. However, the number of closed citations to journal articles within Crossref is also of great significance, since these 322 million closed citations represent 43% of the total.

It is important to note that about one third of these closed citations to journal articles (according to Figure 2) are references to entities published by Elsevier, and that references from within Elsevier’s own publications constitute the largest proportion of these closed citations, since Elsevier is the largest publisher of journal articles. Thus Elsevier’s present refusal to open its article references is contributing significantly to the invisibility of Elsevier’s own publications within the corpus of open citation data that is being increasingly used by the scholarly community for discovery, citation network visualization and bibliometric analysis.

It is also worth mentioning the discrepancy between the citations available in COCI, which comes from the data contained in the open and limited Crossref datasets as of 3 October 2018, and those available within those same Crossref datasets as of 24 January 2019. The most significant difference relates to IEEE. While the citations present in COCI include those from IEEE publications to other entities prior to November 2018 (since in October 2018 its article metadata with references were present within the Crossref limited dataset), in November 2019 this scholarly society decided to close the main part of its Crossref references, and thus from that moment they became unavailable to Crossref Metadata Plus members such as OpenCitations, as highlighted in Figure 3. Thus IEEE citations from articles whose metadata was submitted to Crossref after the date of this switch to closed will no longer be automatically ingested into COCI.

To date, the majority of the citations present in Crossref that are not available in COCI comes from just three publishers: Elsevier, the American Chemical Society and University of Chicago Press (Figure 3). In fact, considering the average value of 18.6 DOI-to-DOI citation links for each citing entity – calculated by dividing the total number of citations in COCI by the number of citing entities in the same dataset – these three publishers are holding more than 214 million DOI-to-DOI citations that could potentially be opened. (The IEEE citation data which was in the Crossref ‘limited’ category as of October 2018 are actually included in COCI, although those from that organization’s more recent publications will no longer be, as mentioned above).

We think it is deeply regrettable and almost incomprehensible that any professional organization, learned society or university press, whose primary mission is to serve the interests of the practitioners, scholars and readers it represents, should choose not open all its publications’ reference lists as a public good, whatever secondary added-value services it chooses to build on top of the citations that those reference lists contain.

CROCI, the Crowdsourced Open Citations Index

The results of the Initiative for Open Citations (I4OC) have been remarkable, since its efforts have led to the liberation of millions of citations in a relatively short time. However, many more citations, the lifeblood of the scholarly communication, are still not available to the general public, as mentioned in the previous section. Some researchers and journal editors, in particular, have recently started to interact with publishers that are not participating in I4OC, in attempts to convince them to release their citation data. Remarkable examples of these activities are the petition promoted by Egon Willighagen (https://tinyurl.com/acs-petition) addressed to the American Chemical Society, and the several unsuccessful requests made to Elsevier by the Editorial Board of the Journal of Informetrics, which eventually resulted in the resignation of the entire Editorial Board on 10 January 2019 in response to Elsevier’s refusal to address their issues (http://www.issi-society.org/media/1380/resignation_final.pdf).

To provide a pragmatic alternative that would permit the harvesting of currently closed citations, so that they could then be made available to the public, we at OpenCitations have created a new OpenCitations Index: CROCI, the Crowdsourced Open Citations Index, into which individuals identified by ORCiD identifiers may deposit citation information that they have a legal right to submit, and within which these submitted citation data will be published under a CC0 public domain waiver to emphasize and ensure their openness for every kind of reuse without limitation. Since citations are statements of fact about relationships between publications (resembling statements of fact about marriages between individual persons), they are not subject to copyright, although their specific textual arrangements within the reference lists of particular publications may be. Thus the citations from which the reference list of an author’s publication has been composed may legally be submitted to CROCI, although the formatted reference list cannot be. Similarly, citations extracted from within an individual’s electronic reference management system and presented in the requested format may be legally submitted to CROCI, irrespective of the original sources of these citations.

To populate CROCI, we ask researchers, authors, editors and publishers to provide us with their citation data organised in a simple four-column CSV file (“citing_id”, “citing_publication_date”, “cited_id”, “cited_publication_date”), where each row depicts a citation from the citing entity (“citing_id”, giving the DOI of the cited entity) published on a certain date (“citing_publication_date”, with the date value expressed in ISO format “yyyy-mm-dd”), to the cited entity (“cited_id”, giving the DOI of the cited entity) published on a certain date (“cited_publication_date”, again with the date value expressed in ISO format “yyyy-mm-dd”). The submitted dataset may contain an individual citation, groups of citations (for example those derived from the reference lists of one or more publications), or entire citation collections. Should any of the submitted citations be already present within CROCI, these duplicates will be automatically detected and ignored.

The date information given for each citation should be as complete as possible, and minimally should be the publication years of the citing and cited entities. However, if such date information is unavailable, we will try to retrieve it automatically using OpenCitations technology already available. DOIs may be expressed in any of a variety of valid alternative formats, e.g. “https://doi.org/10.1038/502295a”, “http://dx.doi.org/10.1038/502295a”, “doi: 10.1038/502295a”, “doi:10.1038/502295a”, or simply “10.1038/502295a”.

An example of such a CVS citations file can be found at https://github.com/opencitations/croci/blob/master/example.csv. As an alternative to submissions in CSV format, contributors can submit the same citation data using the Scholix format (Burton et al., 2017) – an example of such format can be found at https://github.com/opencitations/croci/blob/master/example.scholix .

Submission of such a citation dataset in CSV or Scholix format should be made as a file upload either to Figshare (https://figshare.com) or to Zenodo (https://zenodo.org). For provenance purposes, the ORCID personal identifier of the submitter of these citation data should be explicitly provided in the metadata or in the description of the Figshare/Zenodo object. Once such a citation data file upload has been made, the submitter should inform OpenCitations of this fact by adding an new issue to the GitHub issue tracker of the CROCI repository (https://github.com/opencitations/croci/issues).

OpenCitations will then process each submitted citation dataset and ingest the new citation information into CROCI. CROCI citations will be available at http://opencitations.net/index/croci using an appropriate REST API and SPARQL endpoint, and will additionally be published as periodic data dumps in Figshare, all releases being under CC0 waivers. We propose in future to enable combined searches over all the OpenCitations indexes, including COCI and CROCI.

We are confident that the community will respond positively to this proposal of a simple method by which the number of open citations available to the academic community can be increased, in particular since the data files to be uploaded have a very simple structure and thus should be easy to prepare. In particular, we hope for submissions of citations from within the reference lists of authors’ green OA versions of papers published by Elsevier, IEEE, ACS and UCP, and from publishers not already submitting publication metadata to Crossref, so as to address existing gaps in open citations availability. We look forward to your active engagement in this initiative to further increase the availability of open scholarly citations.

Acknowledgements

The authors would like to thank the SoS Gang (https://sosgang.github.io) for their support, and for having make available a space (https://github.com/sosgang/pushing-open-citations-issi2019) within which to share openly all the scripts and data developed for this study.

Postscript

On the 5th February 2019, after these analyses had been concluded and the text of this paper had been finalized, Crossref published an announcement that DOIs were missing from approximately 11% of its references (https://www.crossref.org/blog/underreporting-of-matched-references-in-crossref-metadata/), because of an historical fault in the manner in which Crossref automatically processes references upon deposit by publishers. This fault means that the absolute numbers of the open and closed DOI-to-DOI citations given in this paper are significantly lower than they should be. While this does not invalidate the comparisons we have reported here, it is clearly regrettable. In April 2019, after Crossref have completed the re-processing of their reference data to include the missing DOIs, we will create an updated version of COCI, and will then recompute and republish the data presented here to include those citations for which Crossref is presently failing to correctly assigned DOIs to the cited entities.

References

Burton, A., Fenner, M., Haak, W. & Manghi, P. (2017). Scholix Metadata Schema for Exchange of Scholarly Communication Links (Version v3). Zenodo. DOI: https://doi.org/10.5281/zenodo.1120265

Heibi, I., Peroni, S. & Shotton, D. (2019). Types, open citations, closed citations, publishers, and participation reports of Crossref entities. Version 1. Zenodo. DOI: https://doi.org/10.5281/zenodo.2558257

OpenCitations (2018). COCI CSV dataset of all the citation data. Version 3. Figshare. DOI: https://doi.org/10.6084/m9.figshare.6741422.v3

Peroni, S., Dutton, A., Gray, T. & Shotton, D. (2015). Setting our bibliographic references free: towards open citation data. Journal of Documentation, 71: 253-277. DOI: https://doi.org/10.1108/JD-12-2013-0166

Peroni, S. & Shotton, D. (2018). Open Citation: Definition. Version 1. Figshare. DOI: https://doi.org/10.6084/m9.figshare.6683855

Schiermeier, Q. (2017). Initiative aims to break science’s citation paywall. Nature News. DOI: https://doi.org/10.1038/nature.2017.21800

Shotton, D. (2013). Open citations. Nature, 502: 295-297. DOI: https://doi.org/10.1038/502295a

Shotton, D. (2018). Funders should mandate open citations. Nature, 553: 129. DOI: https://doi.org/10.1038/d41586-018-00104-7

Sugimoto, C. R., Waltman, L., Larivière, V., van Eck, N. J., Boyack, K. W., Wouters, P. & de Rijcke, S. (2017). Open citations: A letter from the scientometric community to scholarly publishers. ISSI. http://www.issi-society.org/open-citations-letter/ (last visited 26 January 2018)

van Eck, N.J. & Waltman, L. (2010). Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics, 84(2): 523-538. DOI: https://doi.org/10.1007/s11192-009-0146-3

van Eck, N.J., Waltman, L., Larivière, V. & Sugimoto, C. R. (2018). Crossref as a new source of citation data: A comparison with Web of Science and Scopus. CWTS Blog. https://www.cwts.nl/blog?article=n-r2s234 (last visited 26 January 2018)

Citations as First-Class Data Entities: The Open Citation Identifier Resolution Service

Requirements for citations to be treated as first-class data entities

In my introductory blog post, I listed five requirements for the treatment of citations as first-class data entities. The fifth and final of these requirements is that there must be a Web-based identifier resolution service that takes the citation identifier as input and returns a description of the citation.

At the recent PIDapalooza Conference on persistent identifiers, held in Gerona, Spain, I described the Open Citation Identifier Resolution Service, the new resolution service for Open Citation Identifiers created and operated by OpenCitations [1].

In this post, I describe this Open Citation Identifier Resolution Service, which supports the resolution of Open Citation Identifiers not only of the citations documented in the OpenCitations Corpus (OCC), but also of open citations recorded in other bibliographic databases.

What is the Open Citation Identifier Resolution Service

The Open Citation Identifier Resolution Service runs on the OpenCitations server, presenting itself to the user as a web page with the URI http://opencitations.net/oci.

When a user enters a valid OCI and clicks the “Look up citation” button, this activates the resolution service, which, after a brief delay, returns information about the citation itself and about the citing and cited bibliographic resources, as shown in the following screen image (which for clarity omits the provenance data associated with this citation).

This information can optionally be returned to the user in a variety of other formats: RDF/XML, Turtle or JSON-LD.

Clicking on the links provided will return additional metadata held by the OpenCitations Corpus for the citing and the cited documents. In the near future, this service will be integrated with LUCINDA, the forthcoming OCC browse interface, to present this information in a more user-friendly fashion.

Using the Resolution Service with citations in an external resource via a SPARQL endpoint

The Open Citation Identifier Resolution Service currently works for citations between bibliographic resources both within the OpenCitations Corpus and within external bibliographic databases, provided that the external service uses bibliographic resource identifiers having a unique numerical part, and provides a SPARQL endpoint to makes available information about bibliographic resources and the references they contain.

It can therefore resolve OCIs identifying citations within Wikidata, such as oci:01027931310-01022252312, where, as explained in the previous blog post, “010” is the assigned OCC supplier prefix for Wikidata.

Entering this OCI in the Open Citation Identifier Resolution Service pulls live data from the Wikidata SPARQL endpoint and returns the following information about that citation, as shown in the following screen image (which, again, omits for clarity the provenance data associated with that citation):

Clicking on the links provided here returns information about the relevant Wikidata entities.

Citing paper:

Cited paper:

How the Resolution Service works

The bibliographic database supplying the metadata for a particular citation identified by an OCI is specified by the assigned OCC supplier prefix that forms part of the OCI, as described in the previous blog post. Each OCI is thus specific for and unique within a particular bibliographic database.

The resolution service takes the OCI entered into the search box, recognises the supplier prefix specifying the bibliographic database holding the citation information, parses the OCI into the database identifiers for the citing and cited entities, and then sends an appropriate SPARQL query to interrogate the SPARQL endpoint of the relevant database. When that database has returned information about the citation itself and about the citing and cited bibliographic resources, this is displayed to the user as shown in screen images above – or in other RDF formats (Turtle, JSON-LD, RDF/XML) according to the request.

It is important to realize that no other databases are contacted during this resolution process, and that the quality and accuracy of the metadata retrieved by the Open Citation Identifier Resolution Service is the responsibility of the database hosting that citation. The OCI Resolution Service does no more than retrieve this information, and does nothing to address possible errors or omissions in the metadata coming from the hosting database.

Using the Resolution Service with external citations via a REST API

While the resolution service presently works only to retrieve information from bibliographic databases having a SPARQL endpoint, we plan soon to extend this resolution service to work with information supplied by a bibliographic database via a REST API.

Coupled with the ability to create OCIs by numerical conversions of Digital Object Identifiers (DOIs), as explained in the previous blog post, the Open Citation Resolution Service could then be used to pull metadata live from the Crossref REST API for any of the ~350 million Crossref open references in which the cited paper as well as the citing paper has a DOI, and for which an OCI can thus be created.

Watch this space!

References

[1] David Shotton (2018). Citations as first-class data entities. Open Citation Identifiers. Conference presentation. PIDapalooza 2018, Girona, 23-23 January 2018. https://doi.org/10.6084/m9.figshare.5844972

Citations as First-Class Data Entities: Open Citation Identifiers

Requirements for citations to be treated as First-Class Data Entities

In my introductory blog post, I listed five requirements for the treatment of citations as first-class data entities. The fourth of these requirements is that they must be identifiable using a global persistent identifier scheme.

At the recent PIDapalooza Conference on persistent identifiers, held in Girona, Spain, I launched the Open Citation Identifier (abbreviated OCI, in line with DOI), the new persistent identifier for citations [1].

In this post, I describe the Open Citation Identifier scheme, created and operated by OpenCitations, which supports the assignment of Open Citation Identifiers not only to the citations present in the OpenCitations Corpus (OCC) but also to open citations present in other bibliographic databases.

Structure and syntax of the Open Citation Identifier

Each OCI has a simple structure: oci:number-number, where “oci:” is the identifier prefix.

OCIs for citations stored within the OpenCitations Corpus are constructed by combining the OpenCitations Corpus local identifiers for the citing and cited bibliographic resources, separating them with a dash. (For definition of OCC local identifiers, see the OpenCitations Data Model).

For example, oci:2544384-7295288 is a valid OCI for the citation between two papers stored within the OpenCitations Corpus, the first number being the OCC local identifier for the citing bibliographic resource [2], and the second being the OCC local identifier for the cited bibliographic resource [3], these bibliographic resource local identifiers being unique within the OCC. [Note: Supplier prefixes are omitted from OCC local identifiers of bibliographic resources ingested into the OpenCitations Corpus prior to February 2018, but will be included within all OCC local identifiers of bibliographic resources ingested into Corpus after that date.]

OCIs for external resources identifies by numerical identifiers

OCIs can also be created for bibliographic resources described in an external bibliographic database, if they are similarly identified there by identifiers having a unique numerical part. For example, the OCI for the citation that exists between Wikidata resources Q27931310 (the citing resource, [4]) and Q22252312 (the cited resource, [5]) is oci:01027931310–01022252312, where “010” is the assigned OCC supplier prefix for Wikidata. (The colours here and below are added simply for clarity.)

The OCC supplier prefix consist of a positive number (following the pattern “nnn”, where “nnn” is a string of numerals of variable length which includes no zeros), enclosed between two zeros (e.g. “0420”). The list of all assigned OCC supplier prefixes is given at https://github.com/opencitations/oci/blob/master/suppliers.csv.

OCIs for citations between resources identified by DOIs

OCIs can also be created for bibliographic resources described in external bibliographic database such as Crossref or DataCite where they are identified by alphanumeric Digital Object Identifiers (DOIs), rather than purely numerical strings.

To achieve this, 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. For example, using this lockup table, “1” becomes “01”, “2” becomes “02”, “a” becomes “10”, “b” becomes “11”, and “/” becomes “36”. To the resulting number, the appropriate OCC supplier prefix is then added, to clearly identify its provenance.

A citation documented in Crossref exists between the two publications [3] and [6], which are there identified by the DOIs doi:10.1108/jd-12-2013-0166 and doi:10.1371/journal.pcbi.1000361. We can thus create an OCI for this Crossref citation by using numerical representations of the two DOIs. These numerical representations are:

0200101000836191363010263020001036300010606

and

02001030701361924302723102137251211183701000000030601

where the initial “020” in each case is the assigned OCC supplier prefix for Crossref.

From these two numerical representations of DOIs, the OCI for the Crossref citation between these two paper is easily constructed, and is:

oci:0200101000836191363010263020001036300010606-02001030701361924302723102137251211183701000000030601

While this is long for an identifier, it should be remembered that it will be processed computationally, and is not intended for human readability.

In this way, Crossref OCIs can be assigned to all ~350 million open references within Crossref in which the cited paper as well as the citing paper has a DOI [7].

OCIs for the same citation recorded within different databases

If a citation is recorded in more than one bibliographic database, a separate OCI can be created for each instance, each OCI having a distinct supplier prefix and being specific to that database.

Thus, in addition to the Crossref OCI created from DOIs and described above for the citation from [3] to [6], a Wikidata OCI exists for the same citation recorded within Wikidata, having the form oci:01024260641-01021092566.

Upon resolution of an OCI, the Open Citation Identifier Resolution Service will pull metadata only from the database specified by the supplier prefix of the OCI. Details of the Open Citation Identifier Resolution Service are given in the next blog post.

It is important to note that an OCI can only be used to specify a citation between a citing and a cited publication which is actually recorded within a bibliographic database. For this reason, the OCI “oci:7295288-3962641” shown below the second diagram in the introductory blog post to this series is presently invalid. While the OpenCitations Corpus has metadata describing both bibliographic resources [3] and [6], it has not yet ingested the reference list for the first bibliographic resource [3] (which has the OCC local identifier 7295288), having information about it only from a reference within a third paper, with no information about the references [3] itself contains. As a result, at present OCC has no record that a citation actually exists between [3] and the second bibliographic resource [6] (which has the OCC local identifier 3962641).

Representing OCIs in RDF

To permit the description of OCIs in RDF, “oci” has been added as a new member of the class datacite:ResourceIdentifierScheme within the DataCite Ontology.

The resolvable URL for any citation identified by a OCI has the form “https://w3id.org/oc/virtual/ci/nnn-mmm”, where nnn-mmm represents the OCI with its “oci:” prefix removed. Currently, we are able to return the RDF description of all the citations contained in the OpenCitations Corpus and Wikidata. We are working to extend the coverage so as to include other datasets, e.g. Crossref.

References

[2] Armen Yuri Gasparyan, Marlen Yessirkepov et al. (2015). Preserving the integrity of citations and references by all stakeholders of science communication. J. Korean Med. Sci. 30:1545-1552. (English.) https://doi.org/10.3346/jkms.2015.30.11.1545

[3] Silvio Peroni, Alexander Dutton, Tanya Gray and David Shotton (2015). Setting our bibliographic references free: towards open citation data. Journal of Documentation, 71 (2): 253-277. https://doi.org/10.1108/jd-12-2013-0166

[4] Daniel K. Bricker, Eric B. Taylor et al. (2012). A Mitochondrial Pyruvate Carrier Required for Pyruvate Uptake in Yeast, Drosophila, and Humans. Science 337: 96-100.
https://doi.org/10.1126/science.1218099

[5] Douglas Hanahan and Robert A. Weinberg (2011). Hallmarks of cancer: the next generation. Cell 144: 646–674. https://doi.org/10.1016/j.cell.2011.02.013

[6] David Shotton, Katie Portwin, Graham Klyne and Alistair Miles (2009). Adventures in semantic publishing: exemplar semantic enhancement of a research article. PLoS Computational Biology 5: e1000361. http://dx.doi.org/10.1371/journal.pcbi.1000361

[7] Daniel Ecer (2017). Crossref Data Notebook (updated). Available at https://elifesci.org/crossref-data-notebook

Oxford University Press opens its references!

Good news! Today, on January 16th 2018, Oxford University Press (OUP) announced its participation in the Initiative for Open Citations, and requested Crossref to turn on reference sharing for all OUP deposited references from more than half a million publications. Oxford University Press is the largest university press in the world, publishing in 70 languages and 190 countries.

Their announcement is at https://academic.oup.com/journals/pages/announcements_from_oup/oup_joins_I4OC.

OUP now joins the elite band of four university presses that have already made their references open at Crossref in response to the I4OC call (https://i4oc.org/#publishers).

This decision by OUP has been a long time in gestation – see my 2012 post Oxford University Press to support Open Citations – but is no less welcome for that!

Barriers to comprehensive reference availability

Two significant barriers prevent comprehensive reference availability through Crossref.

The first barrier

First, two-thirds of Crossref’s publisher-members, in particular the smaller ones, do not submit references along with the other details of their publications. Many of these published works are of types (e.g. abstracts, editorials and news items) that lack any references. However, while the number of non-submitted references associated with other publications from these publishers is not known, it is likely to be substantial.

Ironically, quite a number of publishers have their Crossref reference status option set to ‘Open’, and yet fail to submit any references!

All publishers who use Crossref DOIs and submit metadata describing their works to Crossref should be strongly encouraged to start submitting associated reference lists if these exist. Crossref have confirmed that it is easy to do, with or without membership of Crossref’s free and beneficial Cited-by Service that provides publishers with statistics on the citations of their own publications. Help can be provided by Crossref Support (support@crossref.org).

The second barrier

The second barrier to full reference availability is created by publishers that submit references to Crossref, but do not presently make them open. Elsevier is by far the largest member of this group, which also includes the American Chemical Society, IEEE and Wolters Kluwer Health.

It is both quick and easy for a publisher to change its preference setting and request that all the references associated with its DOI prefixes are made open – all it requires is an email request to support@crossref.org. But without such a request, the references will be remain in the default ‘Limited’ status.

References that are not associated with Crossref

There are, of course, many scholarly publications, for example preprints in repositories such as arXiv, and journal articles and monographs from small academic publishers in the Humanities, that do not have Digital Object Identifiers issued by Crossref. There are also an increasing number of datasets in repositories such as Dryad that have associated references to the scholarly literature, but whose DOIs are issued by DataCite. None of these submit references to Crossref where they can be made available via the Crossref API, and separate additional measures will be required to capture and share their references with the community.

The new Crossref reference distribution policy

Since 1st January 2018, Crossref has had a new reference distribution policy, described at https://www.crossref.org/reference-distribution/.

There are three possible options for setting the reference distribution preference from which a publisher can choose, these being ‘Closed’, ‘Limited’ and ‘Open”.

If the ‘Closed’ option is chosen, the references will only be used for the Crossref Cited-by service, and are not distributed via any of the other Crossref interfaces.

If the ‘Limited’ option is chosen, the references will additionally be made available to Crossref metadata subscribers who have signed an agreement for the new Crossref Metadata APIs ‘Plus’ service which also came into effect on 1st January 2018. This is the new Crossref default option.

If the ‘Open’ option is chosen, the references will be openly available to anyone using the Crossref APIs.

Publishers will no longer be able to select the reference distribution preference for individual publications on a case-by-case basis, but rather the preference will be set for all publications appearing under a particular DOI prefix, with the new default option being ‘Limited’.

To date, more than 60 scholarly publishers, including most of the major ones, have responded to the Initiative for Open Citations and have instructed Crossref to set their references to open, as describe in a previous post.

Openness of non-Elsevier references

For completeness, this post, also based on analyses performed by Daniel Ecer of eLife (d.ecer@elifesciences.org) on data he downloaded from Crossref in September 2017 (Ecer, 2017), complements the two preceding posts, and details the openness of references from scholarly publishers other than Elsevier.

The main conclusion is that, of the 650,093,489 references stored in Crossref from journal articles published by publishers other than Elsevier, 486,041,671 (74.76%) are open.

The detailed statistics derived from the Crossref data at the time of sampling relating to all publishers except Elsevier are as follows:

Number of works recorded at Crossref from publishers other than Elsevier

Crossref has records of 93,184,372 works with DOIs, of which 69,699,633 (74.80%) are journal articles and 23,484,739 (25.20%) are works that are not journal articles (i.e. book chapters, proceedings articles, datasets, etc.).

Of the 93,184,372 works, 76,795,932 (82.41%) were published by publishers other than Elsevier.

Of the 69,699,144 journal articles, 54,440,761 (78.11%) were in journals with publishers other than Elsevier.

Of the 23,484,739 works that are not journal articles, 22,355,171 (95.19%) were published by publishers other than Elsevier.

Numbers of non-Elsevier works with references

Of all 76,795,932 works with DOIs recorded in Crossref from publishers other than Elsevier, 27,609,963 (35.95%) have accompanying references and 49,185,969 (64.05%) lack references.

Of the 54,440,761 journal articles recorded in Crossref from publishers other than Elsevier, 23,459,805 (43.09%) have accompanying references, and 30,980,956 (56.91%) lack references.

Of the 22,355,171 works that are not journal articles recorded in Crossref from publishers other than Elsevier, 4,150,158 (18.56%) have accompanying references, and 18,205,013 (81.44%) lack references.

Number of non-Elsevier references at Crossref

Of the 1,075,133,743 references stored in Crossref from all works, 732,513,350 (68.13%) are from works published by publishers other than Elsevier.

Of the 956,050,193 references stored in Crossref from journal articles, 650,093,489 (68.00%) are from journals published by publishers other than Elsevier.

Of the 119,083,550 references stored in Crossref from works that are not journal articles, 82,419,861 (69.21%) are from works published by publishers other than Elsevier.

Average numbers of references per non-Elsevier work

The 732,513,350 non-Elsevier references stored in Crossref come from 27,609,963 works of all types with accompanying references, giving an average of 26.53 references per work.

650,093,489 non-Elsevier references come from 23,459,805 non-Elsevier journal articles with accompanying references, giving an average of 27.71 references per journal article.

82,419,861 non-Elsevier references come from 4,150,158 non-Elsevier works with accompanying references that are not journal articles, averaging 19.86 references per work.

Proportion of non-Elsevier works that have open references

Of the 27,598,963 non-Elsevier works of all type documented in Crossref that have accompanying references, 18,228,221 (66.05%) have open references.

Of the 23,459,805 non-Elsevier journal articles documented in Crossref that have accompanying references, 17,072,801 (72.77%) have open references.

Of the 4,139,158 non-Elsevier works documented in Crossref that are not journal articles and that have accompanying references, 1,155,420 (27.91%) have open references.

Proportion of non-Elsevier references that are open

Of the 732,513,350 references stored in Crossref from all works published by publishers other than Elsevier, 523,186,205 (71.42%) are open, and 209,327,145 (28.58%) are not open.

Of the 650,093,489 references stored in Crossref from journal articles published by publishers other than Elsevier, 486,041,671 (74.76%) are open, and 164,051,818 (25.24%) are not open.

Of the 82,419,861 references stored in Crossref from works published by publishers other than Elsevier that are not journal articles, 37,144,534 (45.07%) are open, and 45,275,327 (54.93%) are not open.

Proportion of references which are not open that are published by publishers other than Elsevier

Of the 551,932,682 references from all works stored at Crossref that are not open, 209,327,145 (37.93%) are from works published by publishers other than Elsevier.

Of the 470,008,522 references from journal articles stored at Crossref that are not open, 164,051,818 (34.90%) are from journal articles published by publishers other than Elsevier.

Of the 81,924,160 references from works that are not journal articles stored at Crossref that are not open, 45,275,327 (55.26%) are from works published by publishers other than Elsevier.

Details for all publishers combined, and for Elsevier separately, are given in the two previous posts.

Reference

Ecer, D. (2017). Crossref Data Notebook. Available at https://elifesci.org/crossref-data-notebook

Elsevier references dominate those that are not open at Crossref

Yesterday (November 23rd 2017) I was working with Daniel Ecer of eLife (d.ecer@elifesciences.org) to dig some hard facts out of the analyses he undertook on data he downloaded from Crossref in September 2017 (Ecer, 2017). Because of its dominant position in the scholarly publishing world, in this, the second of two related posts, I report the results for references from works published by Elsevier.

These show that, of all 956,050,193 references from journal articles stored at Crossref, 305,956,704 (32.00%) are from journal articles published by Elsevier, none of which are in the Crossref “Open” category, freely available for others to use.

Put another way, of the 470,008,522 references from journal articles stored at Crossref that are not open, 305,956,704 (65.10%) are from journals published by Elsevier.

On behalf of I4OC, I appeal to Elsevier to join the other major academic publishers and to submit and open all its references without delay.

The detailed statistics derived from the Crossref data at the time of sampling relating to Elsevier publications are as follows:

Number of Elsevier works recorded at Crossref

Of the 93,184,372 works of all types, 16,388,440 (17.59%) were published by Elsevier.

Of the 69,699,633 journal articles, 15,258,872 (21.89%) were published in Elsevier journals.

Of the 23,484,739 works that are not journal articles, 1,129,568 (4.81%) were published by Elsevier.

Numbers of Elsevier works with references

Of all 16,388,440 Elsevier works with DOIs recorded in Crossref, 10,835,273 (66.12%) have accompanying references, and 5,553,167 (33.88) lack references.

Of the 15,258,872 Elsevier journal articles recorded in Crossref, 10,212,958 (66.93%) have accompanying references, and 5,045,914 (33.07%) lack references.

Of the 1,129,568 Elsevier works that are not journal articles recorded in Crossref, 622,315 (55.09%) have accompanying references, and 507,253 (44.91%) lack references.

Number of Elsevier references at Crossref

Of the 1,075,133,743 references stored in Crossref from all works, 342,620,393 (31.87%) are from works published by Elsevier.

Of the 956,050,193 references stored in Crossref from journal articles, 305,956,704 (32.00%) are from journals published by Elsevier.

Of the 119,083,550 references stored in Crossref from works that are not journal articles, 36,663,689 (30.79%) are from works published by Elsevier.

Average numbers of references per Elsevier work

The 342,620,393 Elsevier references stored in Crossref come from 10,835,273 works of all types with accompanying references, giving an average of 31.62 references per work.

305,956,704 Elsevier references come from 10,212,958 Elsevier journal articles with accompanying references, giving an average of 29.96 references per journal article.

36,663,689 Elsevier references come from 622,315 Elsevier works with accompanying references that are not journal articles, averaging 58.92 references per work.

Proportion of Elsevier works that have open references

Of the 10,846,273 Elsevier works of all type documented in Crossref that have accompanying references, 417 (0.0038%) have open references.

Of the 10,212,958 Elsevier journal articles documented in Crossref that have accompanying references, none (0.0000%) have open references.

Of the 633,315 Elsevier works documented in Crossref that are not journal articles and that have accompanying references, 417 (0.0658%) have open references.

Proportion of Elsevier references that are open

Of the 342,620,393 references stored in Crossref from works of all types published by Elsevier, 14,856 (0.0043%) are open, and 342,605,537 (99.9957%) are closed.

Of the 305,956,704 references stored in Crossref from journal articles published by Elsevier, none (0.0000%) are open, 100% being closed.

Of the 36,663,689 references stored in Crossref from works published by Elsevier that are not journal articles, 14,856 (0.0405%) are open, and 36,648,833 (99.9595%) are closed.

Proportion of references which are not open that are published by Elsevier

Of the 551,932,682 references from all works stored at Crossref that are not open, 342,605,537 (62.07%) are from works published by Elsevier.

Of the 470,008,522 references from journal articles stored at Crossref that are not open, 305,956,704 (65.10%) are from journal articles published by Elsevier.

Of the 81,924,160 references from works that are not journal articles stored at Crossref that are not open, 36,648,833 (44.74%) are from works published by Elsevier.

Details for all publishers combined are given in the previous post, and those for all publishers other than Elsevier in the following post.

[Note: As a result of further calculations undertaken by Daniel Ecer on 27th November 2017, which are recorded in his updated Crossref Data Notebook (Ecer, 2017), the figures in this blog have been expanded to show the average number of references for Elsevier works submitting references. At the same time, very minor corrections have been made the total numbers of works in each category, which have not altered the percentages and main conclusions presented in this post.]

Reference

Ecer, D. (2017). Crossref data notebook. Available at https://elifesci.org/crossref-data-notebook

Milestone for I4OC – open references at Crossref exceed 50%

Yesterday (November 23rd 2017) I was working with Daniel Ecer of eLife (d.ecer@elifesciences.org) to dig some hard facts out of the analyses he undertook on data he downloaded from Crossref in September 2017 (Ecer, 2017). In this, the first of two related posts, I report the results for all publishers.

The analyses show that, of the 33,672,763 journal articles documented in Crossref that have accompanying references, 17,072,801 (50.70%) have open references, and of the 956,050,193 references from journal articles stored at Crossref, 486,041,671 (50.84%) are now classified as “Open”, and are freely available for third parties to download and use for any purpose.

This is a significant milestone for the Initiative for Open Citations (I4OC, https://i4oc.org/), which since early 2017 has been campaigning for scholarly publishers to open their reference lists, and a major gain for the world of open scholarship.

The academic community is deeply indebted to all those publishers whose references are now open (https://i4oc.org/#publishers), and to Crossref itself (https://www.crossref.org/), for making these references freely available, providing a tremendous resource for bibliometric analysis.

However, 51.7% of the journal articles recorded in Crossref lack accompanying references, and of the references that are submitted together with the metadata for the remaining journal articles, 49.16% are yet not open.

On behalf of I4OC, I strongly encourage those publishers who are not yet submitting references to Crossref with their article metadata to start to do so, and those other publishers who are submitting references but have not yet made them open to open them without delay, by sending a message requesting this to support@crossref.org.

The detailed statistics derived from the Crossref data at the time of sampling relating to all publishers are as follows:

Works with DOIs documented at Crossref

Numbers of works with references

Of all 93,184,372 works with DOIs recorded in Crossref, 38,445,236 (41.3%) have accompanying references and 54,739,136 (58.7%) lack references.

Of the 69,699,633 journal articles recorded in Crossref, 33,672,763 (48.3%) have accompanying references, and 36,026,381 (51.7%) lack references.

Of the 23,484,739 works that are not journal articles recorded in Crossref, 4,772,473 (20.32%) have accompanying references, and 18,709,979 (79.68%) lack references.

Numbers of references at Crossref

Crossref stores 1,075,133,743 references from all 93,184,372 works.

Of these references, 956,050,193 references (88.92%) are from journal articles, and 119,083,550 references (11.08%) are from works that are not journal articles.

Average numbers of references per work

The 1,075,133,743 references stored in Crossref come from 38,445,236 works of all types with accompanying references, giving an average of 27.97 references per work.

956,050,193 references come from 33,672,763 journal articles with accompanying references, giving an average of 28.39 references per journal article.

119,083,550 references come from 4,772,473 works with accompanying references that are not journal articles, averaging 24.95 references per work.

Proportion of works that have open references

Of the 38,445,236 works of all type documented in Crossref that have accompanying references, 18,228,638 (47.41%) have open references.

Of the 33,672,763 journal articles documented in Crossref that have accompanying references, 17,072,801 (50.70%) have open references.

Of the 4,772,473 works documented in Crossref that are not journal articles and that have accompanying references, 1,155,837 (24.22%) have open references.

Proportion of references that are open

Of the 1,075,133,743 references stored in Crossref from works of all types, 523,201,061 (48.66%) are open, and 551,932,682 (51.34%) are not open.

Of the 956,050,193 references stored in Crossref from journal articles, 486,041,671 (50.84%) are open, and 470,008,522 (49.16%) are not open.

Of the 119,083,550 references stored in Crossref from works that are not journal articles, 37,159,390 (31.20%) are open, and 81,924,160 (68.80%) are not open.

The majority of the references that are not yet open are from works published by Elsevier, as detailed in the next post.

[Note: As a result of further calculations undertaken by Daniel Ecer on 27th November 2017, which are recorded in his updated Crossref Data Notebook (Ecer, 2017), the figures in this blog have been expanded to show the average number of references for works submitting references. At the same time, very minor corrections have been made the total numbers of works in each category, which have not altered the percentages and main conclusions presented in this post.]

Reference

Ecer, D. (2017). Crossref Data Notebook. Available at https://elifesci.org/crossref-data-notebook

Abstract

Introduction

Related works

Indexing citations as first-class data entities

COCI: ingestion workflow, data, and services

Ingestion workflow

Data

Resources and services

Quantifying the use of COCI citation data

Quantitative analysis

Community uptake

Conclusions

Acknowledgements

References

Footnotes

Crowdsourcing open citations with CROCI An analysis of the current status of open citations, and a proposal

Abstract

Introduction

Methods and material

Results

Discussion

CROCI, the Crowdsourced Open Citations Index

Acknowledgements

Postscript

References

Requirements for citations to be treated as first-class data entities

What is the Open Citation Identifier Resolution Service

Using the Resolution Service with citations in an external resource via a SPARQL endpoint

How the Resolution Service works

Using the Resolution Service with external citations via a REST API

References

Requirements for citations to be treated as First-Class Data Entities

Structure and syntax of the Open Citation Identifier

OCIs for external resources identifies by numerical identifiers

OCIs for citations between resources identified by DOIs

OCIs for the same citation recorded within different databases

Representing OCIs in RDF

References

The first barrier

The second barrier

References that are not associated with Crossref

Crowdsourcing open citations with CROCI
An analysis of the current status of open citations, and a proposal