Introduction to Knowledge

Representation and Formal Ontologies

Outline
Ontologies
Data and Knowledge
Do-It-Yourself Ontologies Part I
Do-It-Yourself Ontologies Part II
 Concepts, Symbols, and Things

Humans use symbols (e.g., words) to communicate

Words are mapped to things indirectly through concepts that
denote (refer to) things

Concept

“Jaguar”
Ogden, C. K. & Richards, I. A. 1923. "The Meaning
of Meaning." 8th Ed. New York, Harcourt, Brace &
World, Inc

[Carole Goble, Nigel Shadbolt]

 Concepts, Symbols, and Things
Symbols and concepts are imprecise
◦ The same symbol can stand for multiple things
◦ The same thing can have multiple symbols
◦ Concepts are usually not well-defined

Concept

“Jaguar”
Ogden, C. K. & Richards, I. A. 1923. "The Meaning
of Meaning." 8th Ed. New York, Harcourt, Brace &
World, Inc

[Carole Goble, Nigel Shadbolt]

 Concepts, Symbols, and Things
An ontology attempts to define and relate specific concepts for certain sets of
things via agreed upon symbols

Concept

“Jaguar”
Ogden, C. K. & Richards, I. A. 1923. "The Meaning
of Meaning." 8th Ed. New York, Harcourt, Brace &
World, Inc

[Carole Goble, Nigel Shadbolt]

 Knowledge Representation:
Ontologies
An ontology:
Specifies a theory (a “model”) …
by defining and relating general concepts …
representing features of the world (a domain)

A shared conceptualization (of the “objects, concepts, and other

entities that are assumed to exist” in a particular domain) that is
formally specified (Gruber, 1995)
What are ontologies?
Ontologies are typically created to:

◦ Commit to a definition (a model) of a domain

◦ Explicitly state assumptions concerning the definition

◦ Have a wide scope (be general)

◦ Support exchange and integration of heterogeneous data sources and

applications (more on this later…)
What are ontologies?
Ontologies may be expressed

Informally using natural language (e.g., in philosophy and sometimes biology)

Formally using a mathematical language, e.g., first-order logic (or a fragment)

 What are ontologies?
Ontologies may be expressed

Informally using natural language (e.g., in philosophy and sometimes biology)

Formally using a mathematical language, e.g., first-order logic (or a fragment)

We focus on formal ontologies

◦ To be precise about what the theory proposes

◦ To enable automated processing
 What are ontologies?
Formal ontologies can vary in detail
Controlled Vocabulary
(list of terms)
Simple Thesaurus
(synonyms)
Thesaurus
(broader/narrower terms)

Classification
(class, instance, is-a, maybe part-of)
Classification
(value, cardinality constraints)
Classification
(axioms such as disjoint, union, etc.)
Classification
(general logic constraints)
What are ontologies?
Formal ontologies can vary in detail
Controlled Vocabulary
(list of terms)
Simple Thesaurus
(synonyms)
Thesaurus
(broader/narrower terms)

Expressiveness Classification
(class, instance, is-a, maybe part-of)
Classification
(value, cardinality constraints)
Classification
(axioms such as disjoint, union, etc.)
Classification
(general logic constraints)
What are ontologies?
A conceptualization proposes a theory of the domain of interest
An ontology is a (possibly incomplete) representation of the
conceptualization

set of all theories that can

be expressed in the language

ontology

Conceptualization (of the theory)

[Guarino]
 Class, Instance, and Is-a
“Mulac is a Jaguar”
Jaguar(Mulac)
Jaguar
instance-of
Mulac An instance named “Mulac”
in the set “Jaguar”

Set of things (instances)

denoted by the class Jaguar
Class, Instance, and Is-a
“Every Jaguar is a kind of Animal”
Animal x Jaguar(x) → Animal(x)

is-a
Set of things (instances)
denoted by the class Animal
Jaguar

Set of things (instances)

denoted by the class Jaguar
Defined Properties
Carnivores can be related to other
Animal animals through the “eats” property

is-a eats x,y eats(x, y)  Carnivore(x) →

Animal(y)
Carnivore “if x eats y and x is a carnivore then
y is an animal”
is-a x,y eats(x, y)  Animal(y) →
Jaguar Carnivore(x)
NOTE: Is-a is also a property that relates classes,
but it has a specific, formal interpretation!
Value Restrictions
Animal We are restricting the eats
relationship for Carnivore’s
is-a eats to only Animals
Carnivore
(we are saying that carnivore’s only
is-a eat animals …)

Jaguar
Value Restrictions
Jaguars restrict the eats
relationship to Marsh Deer Animal
x,y eats(x, y)  Jaguar(x) → MarshDeer(y)
eats
Herbivore Carnivore

eats
Marsh Deer Jaguar
Value Restrictions
The formulas …
1. x Carnivore(x) → Animal(x)
2. x Jaguar(x) → Carnivore(x)
3. y Herbivore(y) → Animal(y)
4. y MarshDeer(y) → Herbivore(y)
5. x,y eats(x, y)  Carnivore(x) → Animal(y)
6. x,y eats(x, y)  Jaguar(x) → MarshDeer(y)

If x is a Jaguar, from (2) we know x is a Carnivore

If y is a MarshDeer, from (4) we know y is a Herbivore, and if y is a

Herbivore, from (3) we know y is an Animal
 Value Restrictions
x,y eats(x, y)  Jaguar(x) → ( MarshDeer(y)  Peccary(y) )

Animal
eats
Herbivore Carnivore

eats eats
Marsh Deer Jaguar Peccary
 Value Restrictions
These different representations Animal
propose the same basic
underlying theory eats
JaguarFood Herbivore Carnivore

Marsh Deer
Peccary Jaguar
eats
Cardinality Constraints
1:n
Animal Restrict the number of
participants in a property
is-a eats
Here we say a Carnivore
Carnivore must eat at least one Animal
is-a
Jaguar
 What are ontologies?
Formal ontologies can vary in detail
Controlled Vocabulary
(list of terms)
Simple Thesaurus
(synonyms)
Thesaurus
(broader/narrower terms)

Expressiveness Classification
(class, instance, is-a, maybe part-of)
Classification
(value, cardinality constraints)
Classification
(axioms such as disjoint, union, etc.)
Classification
(general logic constraints)
What are ontologies?
An (informal) ontology of Juice:

◦ Juices are potable liquids made by Juiceries within regions and with
specific vintages

◦ Juices are characterized by the type of grape they are made with,
their color (white, rose, red), their sugar (dry, offdry, or sweet), their
body (light, medium, full), and their flavor (delicate, moderate,
strong)

◦ Sauvignon Blanc, Merlot, Pinot Noir, and Riesling are types of Juices

[OWL Guide]
Exercise
A group of 5, take 10 minutes and try to define a “formal” ontology for
the Juice example

◦ Select three classes

◦ Identify some relationships between them
◦ List any constraints (cardinality or value restrictions) that exist between them
Outline
Preliminaries
Ontologies
Data and Knowledge
Do-It-Yourself Ontologies Part I
Do-It-Yourself Ontologies Part II
Ontologies and Data Management
How do ontologies fit with data management?

◦ Ontologies are kind of similar to conceptual schemas (e.g., E-

R diagrams / DB designs …)
◦ But really they serve different purposes:
◦ Developed independently of a particular application
◦ Given often in a different language
◦ Inherently more general and definitional
◦ Usually not very good schemas
◦ Designed to help support interchange …
 Information Architectures
Ontology
use concepts from
(explicitly or implicitly)

Conceptual Conceptual Design

Model Model Artifact

Schema Schema Schema Schema  Metadata

Data
 Single-world scenarios …
Ecology
Specific
(LTER project xyz)
Project

Scientific (Plant Productivity)

Question

Experiment
(Nitrogen Fertilization over specific plots)
Design

Actual
Actual
Experiments
Actual
Experiments (Different times, plots, samples, etc.)
Experiments

The data/schemas are generally uniform:

Same assumptions, same schemas, …
Data
Collected
 Multi-world scenarios …
Ecology
Project 1 Project 2 Project 3

Integrating across different experiments (e.g., different treatments, plots,

species, methods) requires extra knowledge …
Ontologies can provide the “glue” to resolve differences
◦ To understand the assumptions
◦ To relate the treatments, methods, and so on
Benefits of ontologies
Ontologies are often developed within a community and are
interdisciplinary

Explicitly capture “knowledge” about a domain

◦ Standard vocabulary
◦ Allow “sentences” over that vocabulary

Enable advanced searching techniques

(via reasoning)

Enable exchange and integration

 Benefits of ontologies
Dataset
(Juices by What regional characteristics
regions) produce the best-selling
Juices?

Dataset
(Juice sales) Integrate Analysis

Dataset
(region Integration can be extremely complex
characteristics) due to structural (schema and values)
and semantic (ontological) differences

Ontologies can help!

 Benefits of ontologies
Dataset
(Juices by What regional characteristics
regions) produce the best-selling
Juices?

Dataset
(Juice sales) Integrate Analysis

Dataset
Annotate datasets using ontologies
(region
characteristics)
Map structure (schema) to concepts
Map data to classes/instances
Provides a uniform view of
(various ways to do this…)
disparate sources
Outline
Preliminaries
Ontologies
Data and Knowledge
Do-It-Yourself Ontologies Part I
Do-It-Yourself Ontologies Part II
Constructing Ontologies
In general, creating ontologies is hard

◦ Requires general agreement and understanding of a domain

◦ Requires a clear, concise, and unambiguous definition

◦ May invoke controversy

◦ Is a hard data-modeling problem (complex constraints, broad

domain)
Some strategies …
The goal is to:

◦ Define a vocabulary for your domain (relevant terms)

◦ Define every term in the vocabulary by relating them and giving their
properties …
◦ is-a, instance-of, named properties, constraints, set-based definitions (equivalence, union,
intersection)
 Some strategies …
1. Start with a set of plain-English statements (e.g., “an Ecosystem for our project is …”)

2. List the important terms from the sentences

3. Arrange terms via is-a relationships (and possibly equivalences)

4. Define properties (e.g., spotCount, color; the characteristics)

5. Relate the terms using named relationships (e.g., eats, grownIn)

6. Introduce new terms as needed to relate existing terms

7. Refine the model (simplify, find inconsistencies, etc.)

Exercise
Develop an Ontology
◦ Form groups of 5 people

◦ Pick a project / experimental study design (from your own research)

◦ Define a simplified database schema (two or more tables and their
attributes) for your project/ experimental study design ( 1 mark )

◦ Capture an ontology for your project as box-and-line diagrams

◦ Define the concepts, relationships/properties, and constraints ( 2 marks )
◦ Try to relate the data sets to your ontology … how can someone use your
ontology to understand the assumptions behind the experiment and data
sets? ( 2 marks )
Protégé Demo
Exercise
In your same groups, input your ontology into Protégé

◦ You may need to make some changes …

◦ Be sure to save often!
Some References
Mario Bunge. Treatise on Basic Philosophy, Vol. 3, Ontology I: The Furniture
of the World. D. Reidel Publishing Company, 1977.

Nicola Guarino. Formal ontology and information systems. In Proc. of Formal

Ontology in Information Systems, IOS Press, pp. 3-15, 1998.

Thomas R. Gruber. Toward principles for the design of ontologies used for
knowledge sharing. In Formal Ontology in Conceptual Analysis and
Knowledge Representation, Kluwer Academic Publishers, 1993.

Jeffrey Parsons and Yair Wand. Emancipating instances from the tyranny of
classes in information modeling. In ACM Transactions on Database Systems,
25(2):228-268, 2000.
Some References
Michael Smith, Chris Welty, and Deborah McGuinness. OWL Web
Ontology Language Guide. W3C Proposed Recommendation.
(http://www.w3.org/TR/owl-guide/). Includes Juice Ontology.

Protégé. Stanford Medical Informatics.

http://protege.stanford.edu/index.html. Freely available. Lots of plug-
ins.