on January 22, 2010 by Simon Jupp in Under Review, Comments (2)

Simple Knowledge Organisation System (SKOS)


This article describes the Simple Knowledge Organisation System (SKOS). SKOS is a World Wide Web Consortium (W3C) recommended standard for representing and publishings Knowledge Organisation Systems (KOS) on the Web. A KOS encapsulates a wide range of knowledge artefacts, but can generally be categorised as classifications schemes, thesauri, taxonomies and subject heading systems. KOS have classically been used in the library and informations sciences as a way of indexing large volumes of documents in order to facilitate document retrieval and navigation. The advent of SKOS now enables the construction of KOS that can be used to index documents and share terminologies on the web.

SKOS has been designed in accordance with existing web technologies, the core SKOS data model is defined using vocabulary from OWL, RDFS and RDF, and can be serialised into any of the RDF syntaxes. The core data-model defines the main unit of a SKOS, called a Concept. A SKOS concept can refer to any arbitrary unit of thought and these concepts form the individual terms in the KOS. SKOS provides predicates for specifying relationships between these concepts along with predicates for specifying meat-data such as labels and definitions.

Most KOS typically contain a hierarchical categorisations of concepts, these categories are generally designed with a particular application in mind. One common type of application is a document navigation system where categories are organised in such a way to improve how users browse a document collection. The relationships used to construct these hierarchies are often semantically weak or ambiguous, and are used to indicate a “something to do with” relationship. The hierarchical relationships are generally coined Broader and Narrower, a concept is generally considered narrower if it is somehow more specialised or subsumed by the parent concept. For example, the concept ‘Dog’ would be considered narrower than the concept ‘Animal’.

In additions to relationships between concepts, KOS often contain rich meat-data associated with each term, such as synonyms, hyponyms, spelling and language variations, and natural language definitions. SKOS provides a set of predicated for representing labels, alternate and hidden labels and built in mechanisms from RDF provide support for representing different languages tags. Figure X depicts an abstract representation of some SKOS concepts showing the hierarchical relationships along with some alternate labels for each concept.

Relationship to OWL and Ontology.

Whilst terminologies represented in SKOS closely resemble a class hierarchy from an OWL ontology, it is important to state that both SKOS and OWL have different semantics and are intended for different applications. OWL provides a language for doing knowledge representation and has precise semantics talking about strcit subclass/superclass relationships between classes. SKOS, on the other hand, is not a language for knowledge representation, but rather a language for knowledge organisation. SKOS focuses on terms and how they are organised, the relationships between concepts are semantically weaker. The difference in the semantics of these relationships makes SKOS a useful alternative to OWL in applications scenarios where a rich expressive logic is unnecessary. Figure X attempts to depict where SKOS lies in terms of ontological formality (point to article on formality), it shows how SKOS fits somewhere in-between simple terms lists and strict formal ontology.

SKOS is well placed to bridge the gap between the social tagging culture that has already gained popularity on the web, and the kind of rich metadata and structured knowledge provided by ontologies, that is needed to realise a Semantic Web. There is a close relationship between SKOS and OWL, where SKOS can act as both a sensible intermediate as part of a migratory path to OWL, needs finishing!


  1. dschober

    October 14, 2010 @ 2:42 pm

    typo: strcit should be strict

    “both SKOS and OWL have different semantics and are intended for different applications. ”
    should probably be
    “expressions in SKOS semantics are less formalized and hence can’t be interpreted by Computers and logics reasoners. They are not intended for automatic inference of new knowledge. OWL semantics on the other hand is highly formalized and Computer interpretable allowing for automated reasoning and other advanced AI methods to be applied. “

  2. Robert Stevens

    May 13, 2011 @ 2:37 pm

    Daniel Schober’s comment that one can’t do reasoning on a SKOS vocabualry is wrong. ommonet It is possible to do reasoning on a SKOS artefact, just not very much. One can get trivial entailments like the inverses on SKOS concepts; they are OWL individuals, so when one says that x broaderthan y, the reasoner puts in y narrowerThan x. Similarly, one can exploit transitivity if using DL queries. Similarly, one ccan do defined classes underneath SKOS:concept .

Leave a comment