Information Technology Reference
In-Depth Information
Since RDF is supposed to be an all-purpose knowledge representation system for
the Web, RDF statements themselves can also be described using RDF. RDF itself
has a namespace document,
3
which provides a description of RDF in RDF itself.
In other words, RDF can be meta-modeled using RDF itself, in a similar manner
to the use of reflection in knowledge representation and programming languages
(Smith 1984). For example, the notion of an RDF predicate has its own URI,
4
and is defined there as “the predicate of the subject RDF statement.” The same
holds for almost all RDF constructs, and a conformant RDF processor can derive
from any RDF triple a set of axiomatic triples that define RDF itself, such as
rdf:predicate rdf:type rdf:Property
(all RDF predicates are of the
type property). For a statement like
ex:EiffelTower ex:hasArchitect
ex:Gustave Eiffel
, an agent can infer
ex:hasArchitect rdf:type
rdf:predicate
, which states in RDF that an architect relationship is a predicate
in an RDF triple. However, usually RDF is not hosted according to the Principle
of Self-Description. Use of the media type
application/rdf+xml
is not
consistent usually, and the namespaces URI of specifications like the RDF Syntax
namespace just allow access of to some RDF triples, which is useless to a machine
incapable of understanding RDF in the first place, instead of a more useful RDDL
document (Borden and Bray 2002). A version of RDDL in RDF (Walsh and
Thompson 2007), with an associated GRDDL transform, makes it possible for
Semantic Web agents to follow namespace documents to associated resources
(Connolly 2007).
3.2.4
RDF and the Open World Principle
The Principle of the Open World is the fundamental principle of inference on the
Semantic Web. A relatively simple language for declaring sub-classes and sub-
properties, RDF Schema, abbreviated as RDF(S), was from the beginning part
of the vision of the Semantic Web and developed simultaneously with RDF. Yet
determining how to specify exactly what other triples may be inferred from a given
RDF triple is a non-trivial design problem, since it requires adding an inference
mechanism to a semantic network, which historically in AI featured little or no
inference. Those that do not remember the history of artificial intelligence are bound
to repeat it, and the process of specifying inference in RDF led to an almost complete
repeat of the 'procedural versus declarative' semantics debate. The original RDF
specification defined its inference procedure by natural language and examples. Yet
differing interpretations of the original RDF specification led to decidedly different
inference results, and so incompatible RDF processors. This being unacceptable for
a Web standards organization, the original defender of formal semantics in artificial
Search WWH ::
Custom Search