SKOS-XL for Taxonomies

I recently posted about SKOS (Simple Knowledge Organization System). If you have read anything about SKOS, then you might have come across SKOS-XL (SKOS Extension for Labels) and wondered what that is. The World Wide Web Consortium (W3C) released its recommendations for SKOS and SKOS-XL at the same time in 2009 but chose to make them separate recommendations. One way to see it is that, by separating out SKOS-XL, SKOS is indeed truly “simple.” In the full SKOS reference , SKOS-XL is an appendix.

Extending labels to become resources

“Things, not strings” is a tagline for semantic models, such as SKOS, which emphasize concepts in taxonomies and other knowledge organization systems and not terms or words. Of course, strings of text exist, and when associated with concepts they are called “labels.”  The distinction between a label and the concept that the label describes may seem indistinguishable or perhaps just philosophical. The main difference is that concepts are unique within a taxonomy, but labels are not. A concept may have multiple labels (synonyms or names in different languages), and the same label might apply to different concepts (homographs).

SKOS specifies preferred labels, alternative labels, and hidden labels. Hidden labels can be considered as a type of alternative label that should never be displayed. Alternative labels may display, depending on the front-end application. Preferred labels are what are displayed, especially in hierarchies and facets.

Concepts, as things, have properties or characteristics. Labels do not. But sometimes there are reasons to assign properties to labels, such as to indicate the purpose or use of different labels. In this sense, you would want to turn a string into a thing. More correctly, a thing is called a resource, as described by the Resource Description Framework (RDF) the model upon which SKOS is based. This is what SKOS-XL supports: converting labels to resources. It does this by adding three more elements not found in SKOS: label, label relation, and literal form. It is the label relation in particular that enable the extension to establish a link between label. Further details are in the SKOS-XL recommendation, which I am not going to repeat here.

Use for SKOS-XL

A typical use case for SKOS-XL to assign properties to labels is if you want to have more than one label for different user groups, such as a medical taxonomy for shared medial content to be accessed by both medical professionals and lay people. Medical professionals may prefer a concept labeled Neoplasms, while lay people could call it Cancer. Different user groups could be based in different regions. Although different ISO-code based language labels can be used to distinguish regions in addition to language (such as en-US and en-GB), you may not want to duplicate the vast majority of preferred labels and merely distinguish the few that are actually different.

While SKOS permits multiple alternative labels, aside from hidden labels, there is no way to distinguish their types or purposes in SKOS. You may want to alternative labels support search in one front-end application and not another. You may want to designate official acronyms as distinct from other alternative labels. You may even want to distinguish between different kinds of hidden labels, such as those that should be hidden because they might be pejorative or offensive, and those that you wish to hide only from a type-ahead display because they are near duplicates of other alternative labels and too many alternative labels would clutter up the display.  Finally, there may be alternative labels used by only certain users or in certain regions.

SKOS-XL lets you assign properties or attributes to labels. Assigning the purpose or use of the label is only one possibility, although it is the most common use of SKOS-XL. You may wish to manage more administrative metadata about labels, such as the source or origin of different labels.

Implementing SKOS-XL

The principle of SKOS-XL is not complex, but implementation can be more challenging, and if you are building taxonomies with the SKOS-XL capability, you would want to use taxonomy management software that supports SKOS-XL, such as PoolParty. Taxonomy management software products are quite consistent when it comes to their user interface for supporting the editing of basic SKOS taxonomies, but they are not the same for creating and editing SKOS-XL labels, which is a less common function.

Having properties, such as types, for terms is not new, but required some more innovation in the SKOS model of things (concepts), not strings (terms). It was common for non-SKOS taxonomy/thesaurus management software, which treated different terms with the same meaning as equivalence relationships, to support the customization of relationships, including the equivalence relationship. SKOS-XL ensures that this earlier feature is supported in the current standard, in machine-readable format.

For SKOS-XL to be more widely used and maybe even more elegantly supported requires a great sharing of use cases. I hope the taxonomist community will share their experiences with SKOS-XL, so we can talk about practices and recommendations and not just theory.

Further information:

Leave a Reply