The
Ontology as a Smart, High-Performance
Computer-Based
Information Structure
Purpose
The purpose of this article
is to describe and explain an ontology and how, as a type of computer-based
information structure, it has capabilities that go beyond data
interoperability (i.e., beyond the capabilities of XML schema).
This article will discuss ontology in its intended role of facilitating
communications between applications (i.e., intelligent agents - discussed in
another E-MAPS article) and users with
some level of automatic reasoning.
We will explain how
ontology-based information structures facilitate the transfer of knowledge.
Basic
Concept
The subject of ontology is
complex. We are going to define
the expression twice: once now as
a basic concept and later we’ll define it more fully.
James Hendler defines an
ontology as “…a formal
definition of a body of knowledge.” (1)
This is an excellent
definition because it captures everything clearly without getting too
complicated. We might add,
however, that an ontology is an abstraction that can take the form of a
glossary or other set of concepts or objects that relate to a particular field
of knowledge or collection of objects.
Is this reminiscent of a taxonomy?
Yes, the two are similar. It
is feasible, although not necessarily preferable, to develop an ontology by
first creating a taxonomy of the subject matter and then adding more details
later.
While one cannot “see” an
ontology, just as one cannot “see” a taxonomy, one can visualize it by
looking at a representation or model. In
our article on taxonomy, we introduce them with a representation
of a taxonomy involving areas of the
Now we’ll
model a simple ontology.
We have seen how, over the
past several years, XML has evolved into a robust
technology for the efficient exchange of data.
The supporters of XML and its
schemas include users from commerce, government, the military establishment,
and the World Wide Web. A number of practitioners from these communities
perceive what they characterize as an information glut arising from our own
past successes: near-ubiquitous
computing, the world-wide shift to the World Wide Web, and the rise of an
impressive communications infrastructure to provide ever-increasing bandwidth.
Background:
A Smarter Approach
It is our conviction that
bringing a higher degree of intelligence to our processes – intelligence
derived from the worlds of knowledge representation and artificial
intelligence -- will be of significant benefit for developing smarter systems
capable of greater speed in finding the proverbial needle in the
ever-enlarging haystack.
Our focus on a smarter
approach requires that we clarify some of the differences between knowledge
and information. We will attempt
to address only some of them.
One can readily understand
how knowledge can result from the comparison, synthesis, and evaluation of
information. After processing
information, it is often advantageous to store it in an appropriate repository
(i.e., library, knowledge base) to preclude repeating the process needlessly
and also to make the knowledge available to others.
Explanation
The following definitions,
examples, and exhibits will help explain how an ontology is an
efficient structures for communicating knowledge because of (1) its semantic
advantage (i.e., its ability be exceptionally precise in describing the data
or information it describes); in other words, its ability to describe exact
meaning as opposed to simply providing a match between strings of
characters, and (2) its ability to support rule-based inferential reasoning
exercised by “intelligent agents,” pieces of software also known as
“crawlers”, “bots” or “knowbots”.
A.
Definitions of Ontology
In its original form,
ontology has existed since the time of Aristotle.
Ontology, in antiquity, referred to the study of “being” or
“existence”.
Today, ontology “…
includes data categorization schemes, thesauruses, vocabularies, key-word
lists, and taxonomies. Ontologies
promote semantic and syntactic understanding of data.” (2)
The classic definition of ontology is Gruber’s: “… a specification of a conceptualization”. (3) More loosely speaking, it defines “… the things and rules that exist within a respective domain”. (4)
The following comprise the
essential elements of ontology:
(1)
It is an abstraction, a logical concept.
(2)
Its scope exists within a specific area of knowledge, concepts or
objects.
(3)
The objects or concepts in an ontology have relationships between them;
the relationships are usually but not necessarily hierarchical.
(1) Be used to describe the
components of a given system, area, or domain.
(2) Be used to describe how the
components of a knowledge system relate to one another (i.e., how they
interact).
(3) Be linked together or “chained.”
It is
possible to have a master ontology comprising several constituent ontologies.
(4) Be modeled or
represented in a model as a means of describing them.
(5) Have their logic can be
embedded in software to achieve varying degrees of artificial intelligence.
B.
Visualizing an Ontology
The simplified ontology
modeled below shows entities (things) comprising part of the American
automobile manufacturing industry (domain) and some of the relationships
(rules) between them.
The arrows represent rules or relationships
Both “motor vehicle” and
“corporation” are classes
in the domain of American automobile manufacturing. Classes are groups or
categories having distinct characteristics.
“Automobile” is a sub-class
of the motor vehicle class. Other
sub-classes of “motor vehicle”, although not pictured in the diagram,
could conceivably include “truck”, “tractor”, or “forklift.”
“Automobile manufacturer” is a sub-class of the manufacturer class.
Other sub-classes of class “manufacturer” in this domain could
conceivably include “tire manufacturer” or “glass manufacturer”.
Remember that
“things” in ontology can be abstract or concrete.
Classes and sub-classes are abstract; they don’t actually exist as
anything outside our minds. They
exist only as representations in our minds of something or things that do
exist in reality.
The Ford Taurus LX is
a real entity. One can see and
touch one. It is shown as
an instance
of the automobile sub-class. In
OOP terminology, an instance is a real thing.
It does exist outside our minds. It
is produced by a class or a sub-class. The
act of producing an instance is called instantiation.
We have one final OOP concept
to present - an object.
In real life, an object can be anything.
Generally it’s something that’s real; it can be touched or
perceived in some other physical – versus conceptual -- way.
In OOP, an object is something that’s instantiated from -- or is an
instance of -- a class or sub-class.
Please note the consistency
between this definition and the definition of ontology citing the latter as
“…an abstraction that can take the form of glossary or other set of concepts
or objects that relate to a particular field of knowledge. “
We have discussed classes,
instances, and objects only to enhance one’s understanding of the following
basic concept: ontologies
comprise things --both real and conceptual -- and the relationships between
them within a particular portion of the real world.
Ontologies attempt to simplify the complexities of real life;
they can be thought of as robust models or representations of reality within a
specified scope.
Click here
for a more complete ontology discussion using a wine knowledge base or click here
for a discussion of ontology and NATO’s generic hub.
C.
Purpose/Uses of Ontology-driven Software
The properties,
characteristics, and logic of ontologies may be encapsulated in software
created to accomplish a variety of goals.
The most prominent such uses of ontology-driven software today are in
the fields of knowledge representation, artificial intelligence, and
information technology. These
disciplines have come together with a focus on the Web to produce remarkable
gains in our ability to transfer knowledge.
We wish to parenthetically
note that the following discussion recognizes the large role that the Web and
web-related technology play in the exploitation of ontologies and intelligent
agents. This is not meant to
imply, however, that application of the technology is restricted to the Web.
The technology can be applied in virtually any environment beyond the
web, including mainframe and networked environments.
If the chosen environment is compatible with Java and marked-up text
files, it is compatible with this technology. Note that although Java is not
specifically required, it tends to be the language of choice for these
applications. The reason we
offer web-based examples is because, having started there, this is where the
technology is particularly robust at the present time.
Thanks to the work of the
World Wide Web Consortium, DARPA, and various members of the knowledge
representation/artificial intelligence communities of the European Union, a
phenomenon known as the Semantic Web is emerging.
It should be made clear that the Semantic Web
in the brain child of Tim Berners-Lee, the same visionary who created the
original Web. Just as the
original Web was document-centric with HTML marked-up pages, the semantic web
will be document-centric with semantic
marked-up pages. In the following
text, we refer to such pages as semantic
markup, whether web-based or not.
The difference between
conventional HTML markup and the new semantic markup lies in the language used
to create the markup (or annotation) in the first place.
Originally, the language used was the Hypertext Markup Language, HTML.
It is a very limited language - all
it does is tell how documents should appear when viewed through a browser.
There’s no data involved.
In contrast, semantic markup
empowers documents not only to contain data, but also to contain data in a way
that is far more refined, predictable, and understandable than anything
previously possible. Semantic
markup confers the ability to search and perform other operations with content
– including content outside the web – based on the
meanings of words, or semantics, instead of using simple character
strings having no intrinsic meaning. For
example, instead of searching on “stars” and getting everything from
constellation objects to
It is the ontology that is
directly and uniquely responsible for the logic behind such greatly enhanced
precision, and it is ontology-based markup languages that are directly and
uniquely responsible for the implementation of such logic, or embedding
it, in semantic markup.
D.
How Does Embedding The Logic of Ontology in Semantic Markup Promote
Precision Semantics in the Exchange of Information or Knowledge?
This is the point of
embedding ontology-driven logic into semantic markup – to ensure that the
consumer of information understands precisely what the supplier is providing.
Any potential ambiguity in meaning is minimized because the ontology
has defined its content so precisely.
The ontology facilitates this
process because agreements to its use, as well as detailed understanding of
its precise meanings, are secured in
advance of the exchange of information.
In other words, each party to an exchange – the semantic markup
serving as the source of supply and the intelligent agent as consumer
exercising rule-based inferential reasoning -- knows precisely what the other
is talking about because they have made a prior commitment to use the same
ontology.
The ontological logic and
knowledge embedded in semantic markup is machine-readable
by an intelligent agent as well as by browsers specifically written for the
purpose.
“If
an ontology can be made machine readable, it allows a computer to manipulate the
terms used in the ontology, terms that make sense to users who understand this
information. The computer doesn't understand this information, in any deep
sense of the term, but it manipulates terms that the user understands. This
allows for a form of communication between user and computer, which in turn
enables the creation of software products … that can represent the needs,
preferences, and constraints of the user.“ (1)
The second instance involves
the use of an inference engine employed by an intelligent agent.
When the existing domain knowledge is accessed by the agent, new
knowledge is inferred or deduced after applying the rules or rules of thumb
that reside in the engine.
This is new knowledge. Note - the role of intelligent agents and inference engines is the subject of another paper.
E.
What Are The Principles or Steps Involved in Creating Ontology?
The process of creating
ontology is currently referred to as “ontology engineering”.
Ontologies are engineered by people with appropriate training and
who often perform their work in collaboration with subject-matter experts.
Broadly speaking, we capture
an ontology by defining its purpose and scope, identifying its key components
(objects, concepts, classes) and relationships, identifying terms to refer to
such concepts and relations, and by producing
“unambiguous text definitions” of each.(5)
Noy and McGuiness(6)
recommend the following the seven steps:
1.
Identity the purpose, the domain, and the scope of the ontology.
2.
Consider using a pre-existing ontology, if one exists.
3.
Identify the “important terms in the ontology” (i.e., its properties and relationships
4.
Identify the classes, sub-classes, and their hierarchy.
5.
Identify the properties of the classes, objects or concepts.
6.
Identify the allowed values, value type, etc., of each property.
7.
Create instances of each class or sub-class by defining real objects having
values conforming to the objects’ allowable properties.
The wine knowledge base
ontology discussion (click here)
mentioned previously uses the same seven steps. A reader who wishes to actually develop
an ontology should consider is as a primary starting point.
Conclusion
The thesis of this article is
that today’s computer technology can exploit the capabilities of an ontology
that goes beyond the data interoperability capabilities
of XML.
Embedding the logic of
ontologies in semantic markup -- whether the content is available via the web,
through a corporate intranet, or in a completely closed, offline system
located elsewhere -- is an effective means of storing, processing, and
transferring knowledge.
Knowledge is involved in both
parts of technology’s consumer-supplier relationship.
On the supply side, ontologies embedded in semantic markup capture domain
knowledge by virtue of the construction judgments of ontology engineers in
conjunction with the accumulated wisdom of experienced subject-matter experts
XML schema involves the
storage, processing, and movement of self-described data.
Ontologies created with today’s ontology languages involve the
capture and encapsulation of domain knowledge; when mined by intelligent
agents, new knowledge can be created from the source content so encapsulated.
The following is a partial
list of potential military applications that can exploit ontology-based
systems but could not be accomplished relying upon XML schema technology: