Taxonomic names remain fundamental to linking biodiversity data, but information on these names resides in separate silos. Despite often making their contents available in RDF, records in these taxonomic databases are rarely linked to identifiers in external databases, such as DOIs for publications, or ORCIDs for people. This paper explores how author names in publication databases such as CrossRef and ORCID can be reconciled with author names in a taxonomic database using existing vocabularies and SPARQL queries.
Breaking those silos requires making links between taxonomic names and
other entities of interest, such as the taxonomic literature, specimens, and
taxonomists themselves. Creating these links is currently labour intensive, hence
early efforts at constructing knowledge graphs have either had modest
taxonomic and geographic scope (Ozymandias, [
In this paper I discuss some ways to combine data from taxonomic and publication databases to help seed a biodiversity knowledge graph. In this paper the focus is on plant names, but the ideas apply to names for other taxonomic groups, such as animals and fungi. 2 2.1
The Index of Plant Names Index (IPNI1) is an international register of
published plant names based at the Royal Botanic Gardens, Kew but which has
contributions from the Harvard Gray Index and the Australian Plant Name
Index [
IPNI contains terse citations for the publication of each name in its database. For some records the IPNI curators have added a link to the corresponding page in the Biodiversity Heritage Library (BHL3), and for some recently added records the IPNI web site may give the DOI for a publication, but the majority of IPNI records are not linked to a digital identifier for the publication associated with each name.
As part of ongoing work to match citation strings for taxonomic names to
persistent identifiers for the corresponding publications ([
straightforward task, in that we could treat each work as comprising a set of
pages, and we simply ask which works include the page in the IPNI citation.
However, as yet we don’t have a comprehensive bibliography of life ([
Metadata about the publications themselves publications was harvested from CrossRef5 and ORCID 6, expressed as RDF using terms from the Schema.org7 vocabulary, and added to the knowledge graph. 2.3
I used the same approach to modelling authors employed in Ozymandias [
ScholarlyArticle 10.2307/25065588 ScholarlyArticle
TatianaU.P.Konno https://orcid.org/0000-0001-9036-0912
TatianaU.P.Konno
Although CrossRef does include some author ORCIDs, in many cases it does not. We can match CrossRef and ORCID records using DOIs for publications, and augment CrossRef records with ORCIDs (see Fig. 1). Using a SPARQL query (see Listing 1.1) and a DOI we can match the same author names and order of authorship (role) from CrossRef with those for the same record in ORCID. This query assumes that the author names are identical in the CrossRef and ORCID, an obvious refinement would be to accept approximate matches.
IPNI has a similar, if more complex notion of “roles”. For each taxonomic name there is a “team” of one or more authors, each of whom may have various roles, such as author of the original name, or author of a subsequent version (“combination”) of the name (for example, the change in name that results when a species moves from one genus to another). IPNI records the role of each team member (e.g., whether they are a “publishing author” or a “combination author”) and their position in an ordered list of team members. Hence one approach to matching publication authors (and associated identifiers) to IPNI authors (and their IPNI identifiers) is by matching on roles (Fig. 2). Given this model we can find candidates for matching IPNI team members to publication authors using a SPARQL query (Listing 1.2).
Taxonname
Taxonname tcom:publishedInCitation TatianaUngarettiPaleoKonno
This query finds matches between the first author of the paper with DOI
10.2307 / 25065588 [
In the process of assembling the plant name knowledge graph I have encountered several limitations of the methods outlined here. For example, the order of authorship author in a taxonomic team and in a publication need not be the same. It is not uncommon for an ORCID profile to list only the owner of that profile as the author of a publication, and IPNI only records authors who actually name the plant, which may not be all the authors of corresponding publication. Hence we cannot completely rely on the orderings to make the mapping.
Relatively few taxonomists have ORCID profiles, and these profiles are often incomplete. Indeed of the 483 ORCID profiles currently in the knowledge graph, only 367 include publications, leaving over 100 profiles that cannot be matched to the taxonomic literature using the type of query shown in Fig. 2. The queries described here rely on exact string matching and hence will fail if the source databases do not have the same name for the same person. Currently 6965 IPNI taxonomists in the knowledge graph are linked to a publication with full bibliographic metadata, but only 2101 of those can be matched to one of the authors of that publication. Improving on that matching will require approximate string comparison.
The proposed mapping of authors to identifiers could be also used to
augment existing database records, such as Wikidata8 [
The approach outlined here is merely a first step in fleshing out a knowledge graph of plant names. By connecting names to publications we provide more details on the provenance of those names. Taxonomic publications, especially recent ones, are likely to have details on aspects of the morphology, ecology, and even genomics of the species being considered. By converting text string citations into identifiers such as DOIs, we also open up the possibility of linking the paper
to citing literature, or to additional data from that publication, such as DNA sequences stored in GenBank9, and evolutionary trees stored in TreeBASE10.
There are extensive programs to digitise natural history collections, and the
results of digitisation are being aggregated into global databases such as the
Global Biodiversity Information Facility (GBIF)11. However, data aggregators
have been criticised for not giving attribution and credit to individual researchers
such as taxonomists [
10 https://treebase.org 11 https://gbif.org
PREFIX schema : < http :// schema . org/> SELECT * WHERE {
VALUES ? identifier_value { "10.2307/25065588" } GRAPH < https :// crossref.org > { ? identifier schema : value ? identifier_value . ? work schema : identifier > ? identifier . ? work schema : creator ? role . ? role schema : roleName ? roleName . ? role schema : creator ? creator .
? creator schema : name ? name .
GRAPH < https :// orcid .org > { ? orcid_identifier schema : value ? identifier_value . ? orcid_work schema : identifier ? orcid_identifier . ? orcid_work schema : creator ? orcid_role . ? orcid_role schema : roleName ? orcid_roleName . ? orcid_role schema : creator ? orcid_creator . ? orcid_creator schema : name ? orcid_name .
FILTER (? roleName = ? orcid_roleName) FILTER (? name = ? orcid_name)
Listing 1.1. SPARQL query to match authors in CrossRef and ORCID } } } PREFIX tn: < http :// rs. tdwg. org/ ontology/ voc/ TaxonName#> PREFIX tm: < http :// rs. tdwg. org/ ontology/ voc/ Team #> PREFIX schema : < http :// schema . org/> PREFIX dc: < http :// purl. org/dc / elements/1.1/ > SELECT ? pub_creator ? pub_name ? pub_roleName ? ipni_role_name ? ipni_member_name ? ipni_member ? orcid WHERE {
VALUES ? identifier_value {"10.2307/25065588"} VALUES ? ipni {<urn: lsid : ipni. org: names :77074588 -1 >} ? pub_identifier schema : value ? identifier_value . ? pub_work schema : identifier ? pub_identifier . ? pub_work schema : creator ? pub_role . ? pub_role schema : roleName ? pub_roleName . ? pub_role schema : creator ? pub_creator . ? pub_creator schema : name ? pub_name .
OPTIONAL { ? pub_creator schema : identifier ? person_identifier . ? person_identifier schema : propertyID " orcid " . ? person_identifier schema : value ? orcid .
} } ? ipni tn: authorteam ? ipni_team . ? ipni_team tm : hasMember ? ipni_team_member . ? ipni_team_member tm: role ? ipni_role . ? ipni_team_member tm: index ? pub_roleName . ? ipni_team_member tm: member ? ipni_member . ? ipni_member dc: title ? ipni_member_name .
FILTER (? pub_name = ? ipni_member_name)
Listing 1.2. SPARQL query to match author in IPNI and ORCID