SKOS: Organizing Knowledge with the Simple Knowledge Organization System
In the realm of the Semantic Web, managing how we categorize information is critical for interoperability. The Simple Knowledge Organization System (SKOS), summarized in the W3C Working Group Note known as the SKOS Primer, provides a standardized framework for representing common features found in traditional thesauri and controlled vocabularies.
Unlike traditional systems that focus on terms, SKOS adopts a concept-centric view. In this model, the primitive objects are not the words themselves, but abstract notions—concepts—which are then represented by terms. Each SKOS concept is defined as an RDF (Resource Description Framework) resource, allowing it to be linked and queried across the web.
ไม่มีภาพประกอบ
Key Facts
- Concept-Centric: SKOS treats ideas as abstract entities independent of the labels used to describe them.
- RDF-Based: Every concept is an RDF resource, enabling integration with other semantic web technologies.
- Flexible Labeling: Supports preferred, alternative, and hidden labels in multiple natural languages.
- Hierarchical Structure: Uses broader and narrower relationships to organize concepts.
- Cross-Scheme Mapping: Provides specific properties to link concepts across different vocabularies.
Core Element Categories
SKOS organizes its functionality into several primary categories to ensure a structured approach to knowledge representation.
| Category | Associated Elements |
|---|---|
| Concepts | Concept, ConceptScheme |
| Labels & Notation | prefLabel, altLabel, hiddenLabel, notation |
| Documentation | note, definition, scopeNote, example, historyNote, changeNote, editorialNote |
| Semantic Relations | broader, narrower, related, broaderTransitive, narrowerTransitive, semanticRelation |
| Mapping Properties | broadMatch, narrowMatch, relatedMatch, closeMatch, exactMatch, mappingRelation |
| Collections | Collection, orderedCollection, member, memberList |
Understanding SKOS Concepts and Schemes
At the heart of SKOS is the Concept. These are the fundamental units of thought—ideas, meanings, objects, or events—that underpin knowledge organization. Because concepts are abstract, they remain constant regardless of the language used to label them.
To organize these concepts, SKOS uses a ConceptScheme, which is analogous to a thesaurus or a specialized vocabulary. While a concept can belong to a scheme via the inScheme property, SKOS does not require a concept to be part of a specific scheme, nor does it provide a mechanism to declare a scheme as "complete." Within these schemes, a topConcept serves as one of the highest-level entries in the hierarchy.
Labels and Notations
Labels are Unicode strings associated with concepts. SKOS distinguishes between three types:
- prefLabel: The preferred human-readable term (limited to one per language).
- altLabel: Alternative terms or synonyms.
- hiddenLabel: Terms used for internal processing or search that are not intended for human display.
Additionally, notations are used for classification codes. Unlike labels, notations have specific datatypes (such as integers or dates), making them ideal for alphanumeric codes that are not recognized as standard words.
Documentation and Notes
Documentation properties provide essential context for concepts. All documentation properties are subtypes of skos:note. For instance, the definition property provides a full description of the resource, while scopeNote or example provide further clarity.
These notes are highly flexible; they can be simple strings (literals), resource nodes with their own metadata (such as creator and date), or URIs pointing to external documents.
Semantic Relations and Mapping
SKOS allows users to define how concepts relate to one another, both within a single scheme and across different schemes.
Internal Semantic Relations
Within a scheme, concepts are linked via three primary relationships:
- related: A non-hierarchical associative link between two concepts.
- broader: Indicates a direct hierarchical link where one concept is more general than another.
- narrower: The inverse of the broader relationship.
While broader and narrower are not inherently transitive, SKOS defines broaderTransitive and narrowerTransitive as super-properties to allow for the inference of deeper hierarchical relationships during queries.
Cross-Scheme Mapping
Mapping properties are used exclusively to connect concepts from different schemes. These include broadMatch, narrowMatch, and relatedMatch, which mirror the internal semantic relations. For higher precision, SKOS provides:
- exactMatch: Used when concepts are interchangeable (this property is transitive).
- closeMatch: Used when concepts are similar and occasionally interchangeable, but not identical (this property is not transitive).
Concept Collections
SKOS provides Collection and orderedCollection to group concepts together. These collections can be nested and may be identified by URIs or exist as blank nodes.
It is important to note that a Collection is distinct from a Concept or a ConceptScheme. Semantic relations (like broader) can only be applied to Concepts, not to Collections. To link a concept to others through a collection, individual relations must be defined for each concept within that group.
Frequently Asked Questions
What is the difference between a label and a notation in SKOS?
A label is a human-readable Unicode string (like a name or synonym), whereas a notation is a typed literal (like an integer or date) typically used for classification codes.
Are broader and narrower relations transitive?
No, the basic broader and narrower properties are not declared as transitive. However, SKOS provides broaderTransitive and narrowerTransitive super-properties to enable transitive inference and querying.
When should I use exactMatch versus closeMatch?
Use exactMatch when two concepts from different schemes are interchangeable. Use closeMatch when they are similar and can only sometimes be used interchangeably.
Can a SKOS Collection be a Concept?
No. A SKOS Concept and a Collection are distinct entities; a Concept cannot be a Collection, and a Collection cannot be a Concept.
How does SKOS handle multiple languages?
SKOS supports multiple natural languages by allowing language tags to be attached to labels and documentation properties, ensuring that only one prefLabel exists per language.