Information Technology Reference
In-Depth Information
intelligence, Pat Hayes, oversaw the creation of a declarative, formal semantics for
RDF and RDF(S) in order to give them a principled inference mechanism.
The Open World principle was considered to be a consequence of the lack of
centralized knowledge implied by the decentralized creation of URIs and links as
given by the Principles of Universality and Linking. The parallel to the removal of
centralized link indexes is that on the Semantic Web, “we remove the centralized
concepts of absolute truth, total knowledge, and total provability, and see what
we can do with limited knowledge” (1998c). Hayes argued, in a similar fashion
as he had argued in the original 'procedural versus declarative' semantics debate
in AI, that the Semantic Web should just use standard first-order predicate logic.
Yet while Berners-Lee accepted the need for a logic-based semantics, he argued
against Hayes for the Principle of Open World and monotonicity, and the formal
semantics of RDF was designed to obey the Open World Assumption (Hayes 2002).
The reason for maintaining the Open World Assumption was that adding triples
in a graph merge should never change the meaning of a graph so one could never
retract information by simply adding more triples, or invalidate previously-made
conclusions. This monotonicity is considered key, since otherwise every time an
RDF triple was merged into a graph the interpretation of the graph could change and
so the entire graph might have to be re-interpreted, a potentially computationally
expensive operation. By having a design that allows only monotonic reasoning,
RDF allows interpretations to be changed incrementally in order to scale well in
the potentially unbounded partial information of the Web. Hayes himself eventually
came to agree with Berners-Lee on the issue, noting that reasoning on the Semantic
Web “needs to always take place in a potentially open-ended situation: there is
always the possibility that new information might arise from some other source,
so one is never justified in assuming that one has 'all' the facts about some topic”
(2002).
RDF Schema is on the surface a very simple modeling and inference language
(Brickley and Guha 2004). Due to the Open World assumption, unlike schemas
in relational databases or XML Schemas, RDF Schemas are not prescriptive, but
merely descriptive, and so an agent cannot validate RDF triples as being either
consistent or inconsistent with an RDF Schema (Thompson et al. 2004). They
cannot make the information given by a triple itself change, but only enrich the
description of an existing triple. RDF Schema adds two main features to RDF. First,
RDF(S) provides a notion of class , or a set of resources. Then RDF(S) allows any
resource to be given membership in classes and declare sub-classes (or subsets) of
a class that inherit all the triples created to describe the class. Second, RDF(S) also
allows properties to have sub-properties and for properties to have types for domains
and ranges, such that in a triple the subject is the domain and the object is the
range of a property. Imagine that the property ex:hasArchitect has the range
ex:Person and domain ex:Building . Note that RDF Schemas are not auto-
matically applied to triples even if they are mentioned in a triple, such that for a state-
ment like ex:Eiffel Tower ex:hasArchitect ex:Gustave Eiffel ,
the fact that the domain of ex:hasArchitect is buildings and the range is
people, is not known unless the RDF Schema is automatically imported and
 
Search WWH ::




Custom Search