Languages and standards for the semantic web

Frank van Harmelen, Vrije Universiteit Amsterdam

The essential idea of the Semantic Web is to enrich the existing World Wide Web with machine-interpretable meta-data. Such meta-data will consist of two parts:

• definitions of the terms that are used in the meta-data (for example, in a medical application, we need to define terms like "disease", "patient", "treatment", "drug", "side-effect", etc).
• the actual meta-data that describes the content of web-resources (eg. web-pages, pictures, services) using the terms defined before. (for example, in the medical application, we would expressions such as "this page is about aspirin", "aspirin is a drug", "aspirin is a treatment for headaches", "a side-effect of aspirin is to reduce blood-clotting", etc).

What is obviously needed in such a perspective is a computer language to represent such meta-data, both the definitions of the terms and the expressions using those terms.

The World Wide Web Consortium (W3C, http://w3c.org, the same organisation that has defined HTML, HTTP, XML and all the other standards that keep the web ticking) has defined a set of languages to do exactly this job: define meta-data terms, and express statements about web-resources using those terms.

W3C has defined three languages for this: RDF, RDF Schema, and OWL.

RDF: a simple language for expressing relationships between objects: every statement in RDF is of the form: [thing1] [relationship] [thing2] for example:
http://en.wikipedia.org/wiki/Aspirin
sekt:is-about emtree:aspirin
and emtree:aspirin emtree:preferredTerm "acetylsalicylic acid"

RDF Schema: when stating relations between things in RDF, one uses a specific meta-data terms (such as emtree:aspirin, or sekt:is-about). RDF Schema allows to make a hierarchical organisation of such terms, and to state simple properties of them. For example:
emtree:aspirin rdf:type wikipedia:drug and wikipedia:drug rdfs:subClassOf wikipedia:treatment RDF Schema can state that things are of a certain type, that some types are more general then others, and that some property relates one type to another.

OWL: Clearly, the things we can state about the meta-data vocubulary in RDF Schema are extremely limited: we can state that drugs (like aspirin) have a preferred term (like "acetylsalicylic acid"), but we cannot enforce that each drug has at most one preferred term. We can also not state that drugs and patients are non-overlapping classes (ruling out potential confusions between "baby aspirin" and "baby john"). Similarly, we cannot ecotrin and aspergium are equal to aspirin.

OWL is designed to allow such richer characterisations of a meta-data vocubulary, in which we can say all these things (and more).

Further material

Slide presentations

Slides on the general idea of the Semantic Web, and the role of meta-data in this.
http://www.cs.vu.nl/~frankh/spool/SemWebSlides/SemWeb-tour-Brussels.ppt

A light-weight overview of RDF, RDF Schema and OWL
http://www.cs.vu.nl/~frankh/spool/SemWebSlides/Elsevier04.ppt

More detailed slides on RDF and RDF Schema:
http://www.cs.vu.nl/~frankh/spool/SemWebSlides/RDFv2.ppt

More detailed slides on OWL:
http://www.cs.vu.nl/~frankh/spool/SemWebSlides/OWL.ppt

Text book

Currently the only text-book on the Semantic Web is
A Semantic Web Primer
Grigoris Antoniou and Frank van Harmelen
MIT Press, 40$
ISBN 0-262-01210-3

The book starts with a general introduction to the Semantic Web, and then uses subsequent chapters to travel up the stack of Semantic Web languages: XML, RDF, RDF Schema, OWL, Rules. Two final chapters discuss a dozen use cases for Semantic Web technolgy, and give tips for how to design meta-data vocabularies.

The homepage of the book has additional material such as slides on each chapter: http://www.semanticwebprimer.org/

University course material

We teach a Semantic Web course at the Vrije Universiteit in Amsterdam. Material for this course can be browsed via
http://www.cs.vu.nl/~frankh/webkr.html
It includes slides and lots of on-line background reading material in all grades of technical difficulty, plus further links to other Semantic Web course material on-line.

CLOSE WINDOW TO EXIT