A couple of days ago I wrote:

> 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?

And in response Ben Companjen <[log in to unmask]> wrote the following post, which I think is absolutely wonderful. So wonderful in fact, that I am reposting it with only the tiniest bit of copy editing. I think it is worth a re-read. “Thank you, Ben!”

  While I'm no archivist by training (information systems engineer I am),
  I've learned a thing or two from having to work with EAD and its basis
  for use, ISAD(G) (all citations below are from ISAD(G), 2nd edition). [1] As
  with all information modelling, either inside or outside the Linked Data
  domain, you should take a step back to look at the goal of the
  description. When you have a list of what you want to describe, you can
  start looking for ontologies.

  You probably know this, but I was triggered by "Because many archival
  descriptions are rooted in MARC records, and MODS is easily mapped from
  MARC." to respond. IMO archival descriptions are rooted in rules for
  description, not a specific file format.

  So, when I [see] of (some of) the essences of archival description, I
  think of:

    * "The purpose of archival description is to identify and explain
      the context and content of archival material in order to promote
      its accessibility. This is achieved by creating accurate and
      appropriate representations and by organizing them in accordance
      with predetermined models." (§I.2)

    * "… seven areas of descriptive information:

      1. Identity Statement Area (where essential information is
         conveyed to identify the unit of description)

      2. Context Area (where information is conveyed about the origin
         and custody of the unit of description)

      3. Content and Structure Area (where information is conveyed
         about the subject matter and arrangement of the unit of
         description)

      4. Condition of Access and Use Area (where information is
         conveyed about the availability of the unit of description)

      5. Allied Materials Area (where information is conveyed about
         materials having an important relationship to the unit of
         description)

      6. Note Area (where specialized information and information that
         cannot be accommodated in any of the other areas may be
         conveyed).

      7. Description Control Area (where information is conveyed on
         how, when and by whom the archival description was prepared)."
         (§I.11)

  There is a distinction between the thing being described, and the
  description itself, and both have an important role within the archival
  description. (If anything so far causes confusion with anyone here, I
  misunderstood and accept to be corrected :)) NB: this is one way of
  thinking of descriptions. Incorporating the PROV-ontology would make
  sense for expressing more/other aspects of the provenance of archival
  entities, but I haven't got round to becoming an expert of PROV yet ;)

  ISAD(G) lists 26 "elements that may be combined to constitute the
  description of an archival entity".

  Trying to translate these 'elements', I'd end up with possible a lot
  more than 26 RDFS/OWL properties.

  *Depending on the type of archival entity you can/should of course use
  more specific ontologies.*

  Let me list some properties and related ontologies.

    * Identity statement area

      o Identifiers - The URI, naturally, and other IDs. Could be
        linked using dc(terms):identifier, or mods:identifier, or other
        ontologies. Ideally there is some way of linking the domain of
        the ID to the ID itself, because "box 101" is likely not unique
        in the universe. Perhaps you want to publish a URI strategy
        separately to explain how the URI was assembled/derived.

      o Title - Again DC(terms), MODS, RDA

      o Date(s) - You want properties that have a clear meaning. For
        example, dcterms:created and mods:dateCreated assume it is clear
        what "when the resource was created" means. DC terms are vague, I
        mean general, on purpose. You could create some properties
        `owl:subPropertyOf` dcterms date properties for this. I'd look
        into EDTF for encoding uncertain dates and ranges and BCE dates
        (MODS doesn't support BCE dates).

      o Level of description - What kind of 'documentary unit' does the
        description describe? A whole building's content or one piece of
        paper? I don't know of any ontology with terms "fonds", …,
        "file", "item", but you could say `<http URI> rdf:type <fonds
        class URI>`.

      o Extent and medium - Saying anything about extent and medium
        should possible only happen on the lowest level of description.
        Any higher level extent and medium should be calculated by
        aggregating lower level descriptions. On the lowest level, refer
        to class URIs. A combination of dimensions and material
        {c|sh}ould be a class, e.g. A4 paper 80 grams/square meter.

    * Context area

      o Creator(s) and administrative/biographical history - As ISAD(G)
        refers to ISAAR(CPF) for description of corporate bodies, people,
        and families, this is a perfect example of using existing people-
        and organisation-describing ontologies like FOAF, BIO, ORG, and
        others are useful for separate descriptions of the people and
        organisations involved. You want specific properties to describe
        the roles of these 'agents' in the history of the archival
        entity…

      o Archival history and Immediate source of acquisition or
        transfer - … and you would want them 'here' (of course there is
        no particular order in which these properties are used). PREMIS
        and PROV come to mind first for recording who did what to what,
        (where and?) when and with what result. There are probably some
        ontologies describing possible "events" as RDFS/OWL classes, so
        you could link to those. The immediate source of acquisition or
        transfer may be just another event.

    * Content and structure area

      o Scope and content - Descriptions, keywords, terms from
        authority files about "scope (such as, time periods, geography)
        and content, (such as documentary forms, subject matter,
        administrative processes) … appropriate to the level of
        description.": pretty natural fit for links to SKOS thesauri and
        other ontologies of real-world 'things'. One might think of
        dcterms:subject, dcterms:description, dcterms:temporalCoverage
        etc., but describing *how* exactly such terms relate to the
        archival entity needs more specific properties than "subject" et
        al.

      o Appraisal, destruction and scheduling information - Reasons for
        including things and (possibly) removal of archival entities
        should go very well in rules, and some types of rules go very
        well in ontologies. Making this up as I type: <class of letters
        written by the head of state> rdfs:subClassOf <class of 'things
        to be kept'>. The actual selection and destruction actions could
        be modelled in the same way as other actions are described for
        provenance.

      o Accruals - Whether more content can be expected probably
        depends on other properties of the archival entity, like its
        type(s) and creator(s). I don't know about specific properties to
        record this, but <class of living heads of state archival
        entities> rdfs:subClassOf <class of 'living' archival entities>?
        There are ways of modelling rules for this, like the Rules
        Interchange Format, but the rules may be defined by the archives
        and archivists.

      o System of arrangement - Thinking about this, I tend to think of
        a collection of keywords to describe the arrangement of a
        low-level archival entity like a folder or box: alphabetical, as
        found on deceased's desk. But there is more, of course. Perhaps
        using the Collection Ontology for low levels could help generate
        higher level 'systems of arrangement'.

    * Conditions of access and use area

      o Conditions governing access and Conditions governing
        reproduction - You can describe rights with the Creative Commons
        Rights Expression Language.

      o Language of material - mods:language maybe? Preferably used on
        sub-document level and generated for higher-level descriptions.

      o Physical characteristics / technical requirements - Conditions
        should follow from their respective properties: <class of
        PDF/A-1b files> ..:requiresForReading <class of PDF/A-1b readers>
        and rules that say documents in <class A> are embargoed for 20
        years after creation + a creation date can present enough
        information to the agent to determine dcterms:dateAvailable.

      o Finding aids - As a non-archivist I had some trouble
        understanding the difference between descriptions and finding
        aids and what the exact use of a finding aid was. Also, having
        grown up with search engines, indexes, I think the concept may
        eventually become extinct. I guess you could use foaf:page to
        link a document-like finding aid to the archival entity and
        rdfs:seeAlso to point to machine-actionable related things.

    * Allied materials area

      o Existence and location of originals/copies - PROV can be used
        to link a copy to an original (and how the copy was created
        etc.). `<X> prov:wasDerivedFrom <Y>. <Y> :isAt <AnotherArchive>.`

      o Related units of description / Publication note - Use
        properties that describe the specific relations among archival
        entities. DC Terms has some useful ones, like for citations.
        Related items can be derived from all or selected properties
        automatically too.

    * Notes area

      o Notes - dcterms:description? Unlike a document containing rules
        that needs to be finished at some time, Linked Data has no such
        rule. You can always create a property with a well-defined
        meaning to use for specific information.

    * Description control area

      o Archivist's note / dates of description - Who did what when,
        where, why and how to the description itself. Same as for the
        unit of description itself. This may be a good time to draw a bit
        more attention to the question: *what is a description?* I don't
        have a (/ there is no) final answer, but as The One True Written
        Paper Description from long ago is becoming a set of triples, you
        want to think about it. You could link versions of RDF documents
        using PROV to record this information.

      o Rules and conventions - A link to the rules and conventions for
        description. Could also fit with the PROV provenance.

  No, this is not a list of ontologies to use/explore right away, but I
  hope you (and others) find it helpful, or perhaps even food for
  discussion. Also, have a look at CIDOC-CRM. [2] It has lots of properties.

  [1] ISAD(G) (http://bit.ly/1mmXMmJ) - "This standard provides general
  guidance for the preparation of archival descriptions. It is to be used
  in conjunction with existing national standards or as the basis for the
  development of national standards."

  [2] IDOC-CRM (http://www.cidoc-crm.org) - "The CIDOC Conceptual
  Reference Model (CRM) provides definitions and a formal structure for
  describing the implicit and explicit concepts and relationships used in
  cultural heritage documentation.

  --
  Ben


—
Eric Morgan