Database Reference
In-Depth Information
a computer without analysis by complex natural language processing algorithms,
information is published on the Semantic Web in a structured format that provides
a description of what that information is about. This means that the fundamental
unit of the Semantic Web is an item of data, or a fact, rather than a document. These
facts are brought together to describe things in the real world. Each fact has a basic
structure known as a triple as it comprises three elements: a subject, a predicate,
and an object. These may either be categories of things, such as “River,” individual
things such as the “River Thames,” or just a value. Inevitably, a thing is described
in terms of other things, and this introduces relationships between things, which
parallels the concept of hyperlinks between documents in the traditional Web. For
example, the relationship “flows through” can link the individuals “River Thames”
and “London.” Triples can be interlinked through shared subjects or objects as
shown in the following example:
Subject
Predicate
Object
Des Moines
is part of
Iowa.
Iowa
is part of
The United States of America.
The United States of America
has population
313000000
So that it becomes possible to navigate between facts, in the example we are able to go
from the fact that Des Moines is in Iowa, to learn that Iowa is the in the United States
of America, and then to find out that the United States has a population of 313,000,000.
Many types of predicate or relationship are possible between both categories and
individuals, such as hierarchical (“is a kind of”), mereological (“is a part of”), or spa-
tial (“is adjacent to”), and to specify that an individual is a member of a category (“is
an instance of”). In fact, the author of a Semantic Web description can choose any
relationship the author wishes, thus adding rich meaning to the data on the Semantic
Web. What is particularly interesting is that the item of data may reside in one docu-
ment, or dataset, but be linked to another data item in a different dataset, by any
relationship, but particularly the equivalence relationship “sameAs,” thus instigating
a process of data integration.
Collections of statements about related things in a particular subject area or
domain can be grouped together in what is known as an ontology. An ontology is
therefore more than just a vocabulary, which specifies which terms can be used; or a
taxonomy, which classifies instances into classes. It is a knowledge representation—a
specification of a number of concepts in a domain that are described through the
relationships between them. Some commentators (e.g., in the Semantic Web OWL
standard) (Smith, Welty, and McGuinness, 2004)) point to a further characteristic of
ontologies: reasoning. Statements encoded in OWL can be processed by software
that can reason over them (known as “inference engines,” “rules engines,” or simply
“reasoners”). These reasoners infer logical consequences from a set of axioms or
facts. For example, if we state that “Every Estuary flows into a Sea”
1
and that the
“Thames Estuary is an Estuary,” we can then derive the additional information that
the “Thames Estuary flows into a Sea.” Contrasting this phenomenon to the ordinary
