Conglomerates (EN)#
We want to analyse the following common phenomenon: (1) We have a whole that is more than the sum of its parts. (2) One of the parts gives its name to the whole, other relevant parts are not explicitly mentioned. (3) The whole is described in terms of properties which, strictly speaking, can only meaningfully be applied to individual parts.
Take a didactic example, say “packaged milk”: There’s no more milk in the fridge; we go to the supermarket and buy an item Happy Farmer’s UHT milk 1.5% fat Tetrapack; and voilà, our coffee tastes good again!
What is happening here? We say we buy milk. Actually, however, we do not buy milk, but an everyday commodity “packaged milk” as a whole. (1) This whole is not milk, but a commodity that can be identified by its 13-digit EAN number. And it does not come from cows but from a manufacturing company. (2) The commodity consist of packaging and contents. It is this content “milk” that contributes significantly to the naming of the whole. (3) In the supermarket’s ERP system, this whole is, among others, described by characteristics such as fat content, pasteurisation and packaging manufacturer, which, strictly speaking, make sense only for individual parts of the whole. Let us call this phenomenon a conglomerate: A whole that is described both by the characteristics of its parts and by additional characteristics that do not apply to any of its parts.
Once we have understood the phenomenon of the conglomerate in more detail using the didactic example of packaged milk, we will find it in other contexts of semantic modelling, too. Perhaps the most prominent example is the Dublin Core Initiative. In 1995, at a conference in Dublin/Ohio, it was agreed to describe (web) documents using 15 core elements, including title and author, creator, language, publisher, etc. Around the same time, librarians developed a common understanding that books are made up of entirely different abstract elements: work, expression, manifestation and item, in short: WEMI ([Wiesenmueller 2017] | [WiesenmullerH17])
In our workshop, we identified a Dublin Core document as a conglomerate consisting of WEMI parts. The work as a purely intellectual achievement has a title and an author. The expression expresses an intellectual achievement as a set (an arrangement, a mosaic) of symbols, as “text” in the broadest sense. An important characteristic of expression is the language or the abstract data model: DE, EN, RDF, UML, BPMN, First Order Logic etc. Other modelling paradigms, translations etc. generate new, complementary expressions of the same work. A manifestation encodes an expression technically, it has a format. A printed book, the pdf file or html website with the same text are different manifestations of the same expression; typical characteristics are ISBN, URI or DOI. An item is a tangible book on the shelf or a concrete file on the hard drive that we can access.
We suggest a slightly modified conceptualisation of the WEMI classes: work is a mental image, an idea. An expression is a logical model in the form of a human-readable, structured collection of signs, a “document”. A manifestation encodes an expression in a specific format, a “file”. An item is a concrete copy or download of a manifestation.
This characterisation is merely one of many imaginable ones. Our interpretation implies that ontologically, the WEMI classes are fundamentally different things, and that a central feature of our WEMI conceptualisation lies in a class disjointness similar to owl:AllDisjointClasses. Interestingly enough, as of 2024-08-02, dublincore.org published an OpenWEMI model on its homepage which interprets WEMI as roles and explicitly rejects class disjointness – a decision that restricts compatibility with other important WEMI models, i.e. FRBR from IFLA RDA ([Riva 2018] [])
Familiarity with WEMI allows us to recognise a Dublin Core document as a conglomerate. A single DC record can be transferred to WEMI by creating new associated records for work, expression, manifestation, and item respectively. The way in which the attributes of the original record are mapped to these new records depends on the WEMI ontology to be used, and might look like this (with @prefix dc: "http://purl.org/dc/elements/1.1/"): dc:author and dc:title are shifted to wemi:work; dc:language and dc:creator to wemi:expression; dc:format to wemi:manifestation; and dc:acces_URL to wemi:item.
[Riva 2018] Riva, Pat; Patrick Le Bœuf; Maja Žumer: “IFLA Library Reference Model A Conceptual Model for Bibliographic Information.” IFLA Functional Requirements for Bibliographic Records (FRBR) Review Group, January 2018. https://repository.ifla.org/handle/20.500.14598/40.
[Wiesenmueller 2017] Wiesenmüller, Heidrun und Horny, Silke: Basiswissen RDA. Eine Einführung Für Deutschsprachige Anwender. De Gruyter Saur, Berlin ; Boston, 2., überarbeitete und erweiterte auflage edition, 2017. ISBN 978-3-11-053868-7.