linked-statistics / xkos

A SKOS extension for statistical classifications
35 stars 8 forks source link

sub-/super-ordinate and broader/narrower #16

Closed FranckCo closed 11 years ago

FranckCo commented 12 years ago

Dan is to document the differences between sub/super-ordinate and broader/narrower

FranckCo commented 12 years ago

Here is Dan's contribution :

Broader/Narrower is used in SKOS and defined there to mean a link between two concepts indicating that one is in some way more general ("broader") than the other ("narrower"). The terms are borrowed from ISO 2788 (Documentation - Guidelines for the establishment and development of monolingual thesauri) as indicated in section 8.6.10 and the references (and possibly other places as well).

ISO 2788 defines hierarchical relations and describes "broader" as super-ordination and "narrower" as sub-ordination. Super-ordination and sub-ordination are defined by example along with broader and narrower. There is little formal guidance. Hierarchical relations are said to come in 3 kinds: Generic Partitive Instance These are not defined though they are described, and some examples are given.

Super/Sub-ordinate is used in ISO 704 (Terminology - Principles and methods) and defined in ISO 1087-1 (Terminology - General vocabulary) to refer to concepts in a hierarchical relation. A hierarchical relation is either a generic or a partitive relation. These are defined consistently with how they are described in ISO 2788. In ISO 1087-1, instances are described through the idea of the extension of a concept.

In ISO 704 and ISO 1087-1, the reason for treating hierarchical relations differently than the instance relation is that instances relate concepts with objects, i.e., an object that corresponds to a concept (in a special way that we won't get into here) is said to be in the extension of that concept.

ISO 2788 doesn't make this distinction, because they don't really define concepts. They explicitly say that term and concept are used interchangeably. This means the instance kind of hierarchical relations is not adequately defined.

SKOS does describe a concept as a unit of thought, and this is entirely consistent with ISO 1087-1. SKOS does not appear to make an effort to give examples of broader/narrower relationships. It is not clear at all that the kinds of hierarchical relation described in ISO 2788 are even recognized by the authors of SKOS. Most importantly for us, it is not clear if the authors of SKOS intend for the instance kind of hierarchical relation to be used.

My recommendation is for us to continue to follow ISO 1087-1 for defining hierarchical relations. They are defined there, and they are not defined in any of the other documents listed here. I do not believe that the exclusion of the instance type of hierarchical relation will limit us. There are other more precise ways of handling that situation, namely invoking the extension of a concept.

FranckCo commented 12 years ago

See https://github.com/downloads/FranckCo/SKOSExt/SKOS%20diagram.pdf for a general view of the SKOS/XKOS properties