SKOS: Too Simple?
I gave a keynote at the Extended Semantic Web Conference in Crete last week, speaking on the topic of SKOS, Simple Knowledge Organisation System [1,2]. Towards the end of the talk, I considered the question as to whether SKOS was actually too simple, had insufficient knowledge and a lack of organisation. Personally, I don’t think this is the case. We do still need further evidence and implementation experience, but the growing takeup and usage of SKOS suggests that there is sufficent for many purposes.
During the Q&A, Frank van Harmelen asked a question relating to this point. To paraphrase Frank, the question was essentially “without the additional semantics, what do you get from SKOS that you don’t get from XML?”. At the time, I didn’t give a particular good response to the point (I’m actually pretty terrible at thinking on my feet and answering questions in talks). However, with some reflection, I think that there is enough in there to make it useful and to support interoperation.
Simple put, SKOS essentially allows us to describe concept schemes, collections of concepts that can be used to index and retrieve collections of things. Those concepts can be labelled and can be organised using hierarchical/taxonomic and associative relations. The semantic relationships in SKOS (
skos:related) do not have a formal semantics in the way that OWL subclass hierarchies have a formal semantics. Rather, broader is used to “…to assert that one concept is broader in meaning (i.e. more general) than another…” . A retrieval application using a SKOS vocabulary can then, for example, use the concept taxonomy to expand search terms to cover more potentially related topics. Similarly, the labelling properties (
skos:hiddenLabel) allow the association of labels with concepts, with “terms used a descriptors in indexing systems … represented using [
skos:prefLabel]…” . The presence of alternate labels allows for synonym replacement or again the broadening/narrowing a search processes.
Although these “semantics” (if one could even consider them as semantics) are not expressed in a machine processable form, I believe that the prose in the SKOS Recommendation documents [1,2] provide sufficient descriptions to facilitate interoperation between applications using SKOS vocabularies. If tighter interpretations are required, SKOS allows for extensions to the vocabulary (e.g. through subproperties).
So there is more here than just providing XML tags — we have a vocabulary with an associated description of how that vocabulary should be interpreted. Clearly we need to be careful what assumptions we make when consuming SKOS vocabularies, and it’s perhaps not much, but it’s hopefully enough.
- SKOS Home page
- A. Miles, S. Bechhofer (Eds.) SKOS Simple Knowledge Organization System Reference. W3C Recommendation.
- A. Isaac, E. Summers (Eds.) SKOS Simple Knowledge Organization System Primer. W3C Working Group Note