If you were to select a set of RDF ontologies intended to be used in the linked
data of archival descriptions, then what ontologies would you select?
For simplicity's sake, RDF ontologies are akin to the fields in MARC records or
the entities in EAD/XML files. Articulated more accurately, they are the things
denoting relationships between subjects and objects in RDF triples. In this
light, they are akin to the verbs in all but the most simplistic of sentences.
But if they are akin to verbs, then they bring with them all of the nuance and
subtlety of human written language. And human written language, in order to be
an effective human communications device, comes with two equally important
prerequisites: 1) a writer who can speak to an intended audience, and 2) a
reader with a certain level of intelligence. A writer who does not use the
language of the intended audience speaks to few, and a reader who does not
"bring something to the party" goes away with little understanding. Because the
effectiveness of every writer is not perfect, and because not every reader
comes to the party with a certain level of intelligence, writ!
ten language is imperfect. Similarly, the ontologies of linked data are
imperfect. There are no perfect ontologies nor absolutely correct uses of them.
There are only best practices and common usages.
This being the case, ontologies still need to be selected in order for linked
data to be manifested. What ontologies would you suggest be used when creating
linked data for archival descriptions? Here are a few possibilities, listed in
no priority order:
* Dublin Core Terms - This ontology is rather bibliographic in
nature, and provides a decent framework for describing much of
the content of archival descriptions.
* FOAF - Archival collections often originate from individual
people. Such is the scope of FOAF, and FOAF is used by a number
of other sets of linked data.
* MODS - Because many archival descriptions are rooted in MARC
records, and MODS is easily mapped from MARC.
* Schema.org - This is an up-and-coming ontology heralded by the
600-pound gorillas in the room -- Google, Microsoft, Yahoo, etc.
While the ontology has not been put into practice for very long,
it is growing and wide ranging.
* RDF - This ontology is necessary because linked data is
manifested as... RDF
* RDFS - This ontology may be necessary because the archival
community may be creating some of its own ontologies.
* OWL and SKOS - Both of these ontologies seem to be used to
denote relationships between terms in other ontologies. In this
way they are used to create classification schemes and thesauri.
For example, they allow the implementor to denote "creator" in one
ontology is the same as "author" in another ontology. Or they
allow "country" in one ontology to be denoted as a parent
geographic term for "city" in another ontology.
While some or all of these ontologies may be useful for linked data of archival
descriptions, what might some other ontologies include? (Remember, it is often
"better" to select existing ontologies rather than inventing, unless there is
something distinctly unique about a particular domain.) For example, how about
an ontology denoting times? Or how about one for places? FOAF is good for
people, but what about organizations or institutions?
Inquiring minds would like to know.
—
Eric Morgan
