Semantic Web

Technological Foundations of the Semantic Web

The core idea of the Semantic Web can be boiled down to the handy formulation: things not strings. This basically means that on the Semantic Web, machines are not processing words (strings) but concepts (things) to decipher their meaning in their particular context. This is achieved by embedding these concepts within so-called knowledge models or ontologies. They allow a machine to specify the context in which a certain meaning can claim validity. By doing this a machine is capable of distinguishing between the concept of apple as a fruit, a computer company, or a record label, thus improving the precision of a retrieval system. To achieve this, three methodological steps are obligatory:

• 1.
  Use HTTP-URIs to name things: On the Semantic Web, all things have a unique uniform resource identifier (URI), thus distinguishing ambiguous words from unique concepts. The idea behind this is to use HTTP-URIs—as we know them from the WWW—not just to address but also to name things via so-called namespaces. The HTTP-URI allows machines to dereference things and retrieve associated information, since an HTTP-URI can be linked to other HTTP-URIs according to the principles of hypertext.
• 2.
  Link URIs using RDF: By using RDF (Resource Description Format), HTTP-URIs are semantically linked to graphs. The tiniest graph is called a triple and is represented by the form of subject-predicate-value. Thus it is possible to express simple statements such as Apple (subject) is a (predicate) fruit (value), or William Shakespeare (subject) is author of (predicate) King Lear (value). Specific extensions of RDF—such as RDF Schema—allow linking millions of triples to create giant graphs and build highly expressive data webs. These graphs can be queried with the so-called SPARQL.
• 3.
  Use ontologies to represent context: Ontologies are domain-specific knowledge models that improve the semantic expressivity of RDF. Ontologies can take various forms and differ widely in scope and complexity. An ontology can range from a simple vocabulary (like DublinCore; LOM, for learning object metadata; or SCORM, for sharable content object reference model) that specifies the meaning of concepts within a certain knowledge domain, to highly expressive logic models that specify the relationships between concepts and classes and define logical constraints such as symmetries, transitivities, and inverse relationships. Commonly used standards for the latter case are the Simple Knowledge Organization System (SKOS) or the Web Ontology Language (OWL). From a technological perspective, ontologies lay the basis for a machine to dis-ambiguate, infer, and reason over data, thus improving the linking, integration, coherent navigation, and discovery of knowledge sources.

At this point, it is important to draw a distinction between the Semantic Web and semantic technologies. While the latter denotes a wide range of technologies that are used to algorithmically derive meaning from information; that is, via data mining, natural language processing, reasoning, or tagging, the former should be [Page 636]understood as an endeavor to define standards for representing, linking, and sharing information. Hence, the Semantic Web provides the standards with which semantic data can be represented and published for further processing. Both technological strands complement each and are often symbiotically intertwined.

...