IDEF5
Ontology Description Capture Method
Overview
Historically, ontologies arose from the branch of
philosophy known as metaphysics, which deals with the nature of
reality--of what exists. The traditional goal of ontological inquiry,
in particular, is to divide the world "at its joints,"
to discover those fundamental categories or kinds that define the
objects of the world. So viewed, natural science provides an excellent
example of ontological inquiry. For example, a goal of subatomic
physics is to develop a taxonomy of the most basic kinds of objects
that exist within the physical world (e.g., protons, electrons,
muons). Similarly, the biological sciences seek to categorize and
describe the various kinds of living organisms that populate the
planet.
The natural and abstract worlds of pure science, however,
do not exhaust the applicable domains of ontology. There are vast,
human-designed and human-engineered systems such as manufacturing
plants, businesses, military bases, and universities in which ontological
inquiry is just as relevant and just as important. In these human-created
systems, ontological inquiry is primarily motivated by the need
to understand, design, engineer, and manage such systems effectively.
Consequently, it is useful to adapt the traditional techniques of
ontological inquiry in the natural sciences to these domains as
well.
The IDEF5 method provides a theoretically and empirically
well-grounded method specifically designed to assist in creating,
modifying, and maintaining ontologies. Standardized procedures,
the ability to represent ontology information in an intuitive and
natural form, and higher quality results enabled through IDEF5 application
also serve to reduce the cost of these activities.
Basic Principles of Ontological
Analysis
Ontological analysis is accomplished by examining
the vocabulary that is used to discuss the characteristic objects
and processes that compose the domain, developing rigorous definitions
of the basic terms in that vocabulary, and characterizing the logical
connections among those terms. The product of this analysis, an ontology, is a domain vocabulary complete with a set of precise
definitions, or axioms, that constrain the meanings of the
terms sufficiently to enable consistent interpretation of the data
that use that vocabulary.
An ontology includes a catalog of terms used in a
domain, the rules governing how those terms can be combined to make
valid statements about situations in that domain, and the sanctioned
inferences that can be made when such statements are used in
that domain. In every domain, there are phenomena that the humans
in that domain discriminate as (conceptual or physical) objects,
associations, and situations. Through various language mechanisms,
we associate definite descriptors (e.g., names, noun phrases, etc.)
to those phenomena. In the context of ontology, a relation is a definite descriptor referring to an association in the real
world; a term is a definite descriptor that refers to an
object or situation-like thing in the real world.
In constructing an ontology, we try to catalog the
descriptors (like a data dictionary) and create a model of the domain,
if described with those descriptors. Thus, in building an ontology,
you must perform three tasks: 1) catalog the terms; 2) capture the
constraints that govern how those terms can be used to make descriptive
statements about the domain; and 3) build a model that, when provided
with a specific descriptive statement, can generate the "appropriate"
additional descriptive statements. The expression appropriate
descriptive statements means two things. First, because there
are generally a large number of possible statements that could be
generated, the model generates only the subset that is "useful"
in the context. Second, the descriptive statements that are generated
represent facts or beliefs typically held by an intelligent agent
in the domain who had received the same information. The model is
then said to embody the sanctioned inferences in the domain. It
is also said to "characterize" the behavior of objects
and associations in the domain. Thus, an ontology is similar to
a data-dictionary but includes both a grammar and a model of the
behavior of the domain.
IDEF5 Concepts
The IDEF5 ontology development process consists of
the following five activities.
IDEF5 Ontology Languages
Supporting the ontology development process are IDEF5’s
ontology languages. There are two such languages: the IDEF5 schematic
language and the IDEF5 elaboration language. The schematic language
is a graphical language, specifically tailored to enable domain
experts to express the most common forms of ontological information
(see Figure 1). This enables average users both to input the basic
information needed for a first-cut ontology and to augment or revise
existing ontologies with new information. The other language is
the IDEF5 elaboration language, a structured textual language that
allows detailed characterization of the elements in the ontology. 
Figure
1: Basic IDEF5 Schematic Language Symbols
Various diagram types, or schematics, can be constructed
in the IDEF5 Schematic Language. The purpose of these schematics,
like that of any representation, is to represent information visually.
Thus, semantic rules must be provided for interpreting every possible
schematic. These rules are provided by outlining the rules for interpreting
the most basic constructs of the language, then applying them recursively
to more complex constructs.
However, the character of the semantics for the Schematic
Language differs from the character of the semantics for other graphical
languages. Specifically, each basic schematic is provided only with
a default semantics that can be overridden in the Elaboration Language.
The reason for this is that the chief purpose of the Schematic Language
is to serve as an aid for the construction of ontologies; they are
not the primary representational medium for storing them. That task
falls to the Elaboration Language. The Schematic Language is, however,
useful for constructing first-cut ontologies in which the central
concern is to record, in a rough way, the basic elements that exist
in a domain, their characteristic properties, and the salient relations
that can be obtained among objects of those kinds and among the
kinds themselves. Consequently, the basic constructs of the Schematic
Language are designed specifically to capture this type of information.
IDEF5 Schemantic Types
Certain relations predominate when people express
their knowledge about a domain; because of their prominence and
importance, these relations are included explicitly in the IDEF5
language. There are four primary schematic types derived from the
basic IDEF5 Schematic Language which can be used to capture ontology
information directly in a form that is intuitive to the domain expert.
Classification Schematics
Classification schematics provide mechanisms for
humans to organize knowledge into logical taxonomies. Of particular
merit are two types of classification: description subsumption and natural kind classification. In description subsumption,
the defining properties of the "top-level" kind K in
the classification, as well as those of all its subkinds, constitute
rigorous necessary and sufficient conditions for membership in
those kinds. Additionally, the defining properties of all the
subkinds are "subsumed" by the defining properties of
K in the sense that the defining properties of each kind entail
the defining properties of K; the defining properties of K constitute
a more general concept. Conversely, natural kind classification
does not assume there are rigorously identifiable necessary and
sufficient conditions for membership in the top-level kind K.
Nonetheless, there are some underlying structural properties of
its instances that, when specialized in various ways, yield the
subkinds of K. The difference between the two types of classification
is illustrated in the example below.
Figure 2: Different Types of Classification
Composition Schematics
Composition schematics serve as mechanisms to represent
graphically the "part-of" relation that is so common among
components of an ontology. In particular, this capability enables
users to express facts about the composition of a given kind of
object. For example, one might want to represent the component structure
for a certain kind of ballpoint pen.
Figure 3: Composition Schematic for the Kind Ballpoint Pen
As the schematic in the figure shows, the ballpoint
pen in the domain in question has both an upper body and a lower
body, and that the former consists of a button, a retraction mechanism,
and an upper barrel while the latter consists of a lower barrel
and a cartridge, which in turn consists of a spring and an ink supply.
Relation Schematics
Relation schematics allow ontology developers to visualize
and understand relations among kinds in a domain, and can also be
used to capture and display relations between first-order relations.
The motivation for developing this capability is that people often
describe and discover new concepts in terms of existing concepts.
This means of creating and defining new concepts is consistent with
Ausubel’s theory of learning, wherein learning often occurs
by subsuming new information under more general, more inclusive
concepts (Novak & Gowin, 1984; Sarris, 1992). Based on this
hypothesis, a natural way to describe a new (or poorly understood)
relation is to connect it to a relation that is already well understood
and, more generally, to categorize its place in a "conceptual
space" of other relations. The IDEF5 relation library (included
as an appendix in the IDEF5 Ontology Capture Method Report)
provides a baseline reference to help users discover and characterize
relations.
Object State Schematics
Because there is no clean division between information
about kinds and states and information about processes, the IDEF5
schematic language enables modelers to express fairly detailed object-centered
process information (i.e., information about kinds of objects and
the various states they can be in relative to certain processes).
Diagrams built from these constructs are known as Object-State Schematics.
Two types of changes can be observed in the objects
undergoing processes: change in kind and change in state. There
is no formal difference between these two types of change: objects
of a given kind K that are in a certain state can simply
be regarded as constituting a subkind of K. For formal purposes,
for example, warm water can be regarded as a subkind of water.
However, it is useful to distinguish the two in the schematic language
to indicate explicitly the kind of thing that is in a certain state.
This is done using colon notation (e.g., kind:state). For
example, warm water will be indicated by the label water:warm,
frozen water by water:frozen, and so on. The IDEF5 schematic
language allows modelers to visually represent changes in an object's
kind or state as well as the processes that bring about such changes.
Figure
4: Example Object State Transition Schematic
Summary
The nature of any domain is revealed through three
elements: 1) the vocabulary used to discuss the characteristic objects
and processes comprised in the domain, 2) rigorous definitions of
the basic terms in that vocabulary, and 3) characterization of the
logical connections between those terms. An ontology is a domain
vocabulary complete with a set of precise definitions, or axioms,
that constrain the meanings of the terms in that vocabulary sufficiently
to enable consistent interpretation of data. The IDEF5 method provides
a structured technique, by which a domain expert can effectively
develop and maintain usable, accurate domain ontologies. The IDEF5
method is used to construct ontologies by capturing assertions about
real-world objects, their properties, and their interrelationships.
References
Novak, J., and Gowin, D. B. (1984). Learning How
to Learn. Cambridge: Cambridge University Press.
Sarris, A. K. (1992). "Needs Analysis and Requirements
Document: Integration Toolkit and Methods, Corporate Data Integration
Tools." MANTECH Report WL-TR-92-8027. WPAFB, OH. |
