National Species List - Semantic Web Model

The NSL’s Semantic Web Model comprises the following parts:

The following terms are used in this document

The following elements are used in this document’s Conceptual and Logical model diagrams:

Namespaces for model, vocabulary and validator elements are assigned the following prefixes:

Where example data is given below and dummy values are used, either for modelling elements - classes or predicates - or instances of them - particular persons or names etc. - the prefix ex: is used to indicate that this is example/dummy data only.

This model includes a Taxon class which represents "A group of organisms considered by taxonomists to form a homogeneous unit" and links to that a Taxon Name class which represents "A name applied to Taxa by people at a point in time". Instances of Taxon Name are then modelled to have been used over time through a Taxon Name Usage class which may cite Bibliographic Reference class instances that indicate Creative Work instances which represent the papers/articles/books etc. that provide the taxon name and sometimes describe the taxon. Taxon Name Usage class instances may also be cited and instances of this are recorded as Cited Usage class instances.

An informal overview of this model is given in Figure 3 below.

This model exists within both a national and international domain context - other models within the general area of biodiversity, species and bibliographic referencing - and a technical context of expressing and publishing Semantic Web models.

The auNLS is a stand-alone system, meaning it is not dependent on any other data sources for its interpretation or use, however its content is used by many other systems such as databases of biodiversity observations, in particular the:

These two systems have their own models which necessarily relate to this model.

The BDR dataset uses the ABIS model which contains uses the same Taxon and Taxon Name classes as defined in this model as well as the Bibliographic Reference and Creative Work classes and cites & references predicates this model also reuses for general-purpose modelling. The use of the Taxon and Taxon Name classes by ABIS is shown in the following figure.

This information is repeated in the Mapping Section of this document.

The Atlas of Living Australia data model is, or is becoming, the new GBIF New Data Model which declares a Taxon class which is the same as this model’s Taxon class and which declares an Identification class which is similar to this model’s Taxon Name class.

A Semantic Web form of the GBIF New Data Model is expected in 2024 and this auNSL/GBIF NDM relation will be updated then. For now, the mapping from this model to the GBIF NDM is given in the Mappings Section.

The Semantic Web is the concept of a machine-readable, and perhaps more ambitiously, a machine-understandable Internet.

While the original coining of the phrase in 2001 described "an expected evolution of the existing Web to a Semantic Web" and by 2006 well-known Internet pioneers lamented it as being unrealised ref, in the 2020s, there are vast amounts of data on the Internet encoded in the form expected, in 2001, to be machine-understandable: RDF.

In addition to RDF data on the Internet, there are many other machine-understandable data encodings and originally the NSL was, and still is, encoded in another one, GraphQL. What those other forms of data encoding lack is the very granular, universal, identification of data elements and relationships that RDF has which make the meaning of all data elements in RDF explicit, leaving nothing ambiguous.

By choosing the express the NSL’s data model in RDF, a commitment is being made to expressing all the elements of the model in a form that can be understood without implicit knowledge of the NSL or anything else, since all RDF objects and their definitions are dereferencable - their definitions and details may be obtained from the Internet.

The Semantic Web, in the 2020s, relies on a few simple data models, such as RDF & RDFS for its basic data structures and on OWL, the Web Ontology Language for its modelling paradigm. In addition to these fundamental technical models, there are a series of well-known and often standardised generic domain models that are widely used by many datasets claiming to be part of the Semantic Web. For example, the Provenance Ontology is "…​about entities, activities, and people involved in producing a piece of data or thing…​" and is the Semantic Web’s go-to model for process flow modelling. GeoSPARQL is the Semantic Web’s most widely use spatial data model and schema.org.

There is value in reusing well-known models is for human understanding - it’s much easier to understand a new model if it is linked to things you already know - but also in model quality - many well-known modes are well-known because they are good and widely used.

Here, expressing the NSL’s data model as a Semantic Web model means not only modelling its concepts in OWL using RDF but also reusing, or at least deriving from and aligning to, these well-known models.

While is no definitive set of these well-known models and the set will surely change over time (in the 202s, schema.org is increasingly being used in favour of the earlier and smaller Dublin Core Terms), this model has derived from and extended are currently in wide use and some have been for a decade or are well-known within the taxonomy domain.

The next sections describes well-known model ruse by this model in formal terms and the Mappings section describes this model’s mappings to a number of Semantic Web and similar models that aren’t directly.

A 'Supermodel' is an overarching or enterprise model. The implementation of a focus model within a Supermodel context formally sets the broader context for the model by identifying and typing the other models that the focus model relates to and by defining the relations also.

The general picture of a Supermodel is that it contains:

Modelling Paradigm

Model formalisms

Model technology

In graph modelling, as per RDF and OWL, objects are associate with binary relations which are unqualified. For example, a Creative Work, such as a book, might be attributed to multiple Person instances like this:

In this example, we have no knowledge of what the role of ex:person-a or ex:person-b was/is with respect to the book ex:book-001.

We can use specialised predicates to indicate roles. Building on the example above we might have:

Now we can see the different roles played by the persons - "author" and "editor" - but this is limied to a defined set of predicates.

Instead, we can use the qualified relation pattern (see ref) that interposes a node (data object) between the two related objects - a book and a person - and attached qualifying information to that node. Using this pattern, the above example could be represented like this:

Here, the same information is conveyed as in the previous example however the roles - "author" & "editor" - can be defined in an expandable, or even multiple, vocabularies of terms and do not need to be drawn from a ontology containing predicates. This make the expansion and even alteration of roles much easier.

For a TaxonName object, multiple usages - instances of Usage - are likely to exist and it is important to indicate which of them is the "accepted use". If no qualified relations are used, this model’s predicate nsl:isCitedBy can be used to relate the objects, but they cannot be differentiated, like this:

If the qualified relation pattern is used, the role of the Usage - "accepted" or otherwise - can be seen:

This qualified relation pattern representation of name usage can include temporal details like this:

In the above example, two Usage instances are indicated as having the role of "accepted", but they are further qualified with temporal extents meaning that there has not been more than one "accepted" use here.

Many models - Semantic Web models and others - have patterns for indicating referencing or citation. For example, schema.org includes a predicate `citation - that can be used to associate a citing and a cited object, like this:

This simple form of citation does not cater for qualification of the citation (see the Qualified Relations section above) or for the indication or part citation: which part of Frankenstein was cited.

Detailed citation models handle these issues and the Citation Typing Ontology (CiTO) model which interposes a Citation object between the citing and a cited objects, following the Qualified Relations pattern to which qualifying details can be related, like this:

Here ex:highschool-essay-x cited ex:frankenstein in the manner of a critique - cito:critiques.

In this model, we adopt the CITO cited/citing pattern with renamed predicates of cito:hasCitingEntity → nsl:citing and cito:hasCitedEntity → nsl:cited and make an equivalence to the unqualified citation pattern of schema.org whereby the predicate path sdo:citation == (inverse of) cito:hasCitingEntity / cito:hasCitedEntity, as per the following figure:

The representation of citation of specific parts of a cited work are not provided for directly by CITO but are partly support by the use of in-text reference pointers in the CITO sibling ontology C4O able to be indicated. Another CITO sibling ontology, DOCO, provides for detailed document part modelling which allows citations to indicate a part of a document (a Creative Work) but only if that document has been decomposed into addressable parts.

The reasonably well-known (in ontology circles) BIBO ontology for "expressing citations and bibliographic references" allows for standard scholarly citation element representation: page numbers, journal volumes and so on.

Using the altered citation pattern, described above, as well as model elements from BIBO, we can model NSL Usage instances like this:

The citing and cited objects may vary: TaxonName instances but also other Usage instances may be cited.

The background models in this model are all those standard or common ontologies reused by this one. The main background models and their use by this model are given in the table below.

The relations of this model to the Background Models it inherits from are given in Profiles Vocabulary [PROF] terms in a formal "profile definition". That definition is related here in human-readable form and in machine-readable form (RDF) at:

TODO: write up profile definition

The BDR dataset uses the ABIS model which contains uses the same Taxon and Taxon Name classes as defined in this model as well as the Bibliographic Reference and Creative Work classes and cites & references predicates this model also reuses for general-purpose modelling. The use of the Taxon and Taxon Name classes by ABIS is shown in the following figure.

This information is repeated in the Related Domain Models subsection of the Introduction Section of this document.