Ontologies for Information

Formal versus Material
Ontologies for Information
Systems Interoperation in the
Semantic Web

Work in progress

R. Colomb

School of ITEE, The University of
Queensland

20 August, 2002

Outline

•
Ontologies, upper ontologies,and semantic
heterogeneity

•
Sample of upper ontology efforts

•
Semantic heterogeneity

•
The synthetic a priori of Kant

•
Application to the ontologies

•
So what?

Ontologies, upper ontologies,and
semantic heterogeneity

•
Application
-
specific
-

SIC, SNOMED

•
Upper ontologies application
-
independent

•
…what we now refer to as philosophical ontology has
sought the definitive and exhaustive classification of
entities in all spheres of being … including the types of
relations by which entities are tied together

•
Heterogeneity
-

enemy target example

Sample of upper ontology efforts

•
Cyc

•
SUMO

•
OntoClean

•
GOL

•
Bunge
-
Wand
-
Weber (BWW)

•
WordNet


Cyc

•
Transaction
-

The collection of actions performed by two or
more agents cooperating (willingly) under some agreement
wherein each agent performs actions in exchange for the
actions of the other(s). Note that a case of attack
-
and
-
counterattack in warfare is
not

a Transaction; nor is fortuitous
cooperation without agreement (e.g. where a group of investors
who, unknown to each other, all buy the same stock almost at
once, thereby driving up its price). For transactions involving
an exchange of user rights (to goods and/or money) between
agents, see the specialization of ExchangeOfUserRights

•
Subtype of PurposefulAction, CooperativeEvent


SUMO

•
Transaction

-

The subclass of ChangeOfPossession where something
is exchanged for something else.
subclass of

ChangeOfPossession

•
ChangeOfPossession

-

The Class of Processes where ownership of
something is transferred from one Agent to another.
subclass of

SocialInteraction

•
SocialInteraction

-

The subclass of IntentionalProcess that involves
interactions between CognitiveAgents.
subclass of

IntentionalProcess

•
IntentionalProcess
-

A Process that is deliberately set in motion by a
CognitiveAgent.
subclass of

Process

OntoClean

•
Quality

•
Quality Region

•
Aggregate

–
Amount of matter

–
Arbitrary Collection

•
Object

–

Physical Object

•
Body

•
Ordinary Object

–
Mental Object


•
Feature

–
Relevant Part

–
Place

•
Occurrence

–
State

–
Process

–
Accomplishment


GOL

•
Entity

–
Set

»
Extension

–
Urelement

»
Individual

»

Chronoid (temporal duration)

»

Topoid (spatial region)

»

Substance

»

Moment

»

Quality

»

Relational Moment

»
Universal

»

Relational Universal


Bunge
-
Wand
-
Weber (BWW)

•
Thing

•
Property and Attribute

•
State of a Thing
: at a point in time, the attributes of a thing
have values.

•
Event
: change of state in a thing.

•
History of a Thing
: a sequence of events in a thing.

•
Type/Class and Subtype/Subclass

•
Composite thing:

is composed of (made up of) things other
than itself. Things in the composite are
part
-
of

the
composite.


WordNet

•
Abstraction

•
Act (human)

•
Entity

•
Event

•
Group

•
Phenomenon

•
Possession

•
Psychological Feature

•
State

Semantic heterogeneity

•
Structural

–
Can be overcome with more or less elaborate views

•
vs Fundamental

–
The words mean something a little different in the two
systems

Transaction examples

•
the interaction with Amazon.com resulting in the placing of an
order and the supply of credit card details.

•
the subsequent packing and shipping of the order, its receipt by
the purchaser in good condition, and the acceptance of the
credit card charge by Visa.

•
the interaction with Medline resulting in the placing of a query
and the return of a collection of abstracts

•
the borrowing and ultimate return of a book by the University
of Queensland library from the University of Sydney library
(interlibrary loan), on behalf of an academic (who must also
borrow and return the book from the University of Queensland
library).

•
the interaction between the 2002 Salt Lake City Winter
Olympics results processing agent and the agents responsible
for the maintenance of results on multiple web sites ultimately
completing with the information that the medal results for ice
hockey have been recorded on all sites.


Searle’s Institutional Facts

•
Brute fact X counts as institutional fact Y in
context C

–
Eg Marriage, naming, buying

•
Result of speech act

•
Information systems concerned mostly with
institutional facts

•
Institutional facts need background for
interpretation

Background

–
Searle: The literal meaning of any sentence can
only determine its truth conditions or other
conditions of satisfaction against a background
of capacities, dispositions, know
-
how, etc.,
which are not themselves a part of the semantic
content of the sentence.

•
Eg our expectation of the behaviour of objects in
our environment, and of how various kinds of
situations are supposed to develop.

Institutional facts immanent

•
Nature of institutions means new
institutional facts easily created

–
eg UQITEE
-
funded travel

•
Transcendent system (rules of chess) give
global shape

–
Permit ontology of openings, end games, etc

–
Rule change part of game makes system
immanent, lose ontology.

Institutional Facts Immanent

•
Only local warrant needed for creation of
new institutional fact

•
Institutional facts are in complex contexts,
but ultimately immanent

•
So no reason to expect an a priori ontology

Synthetic a priori

•
We can know a priori only how we
represent knowledge, not what we can know

Synthetic a priori

•
Space

is the form of all appearances of
outer sense, i.e.. the subject condition of
sensibility, under which alone outer
intuition is possible for us

–
Includes identification of objects

Synthetic a priori

•
Time is the form of inner sense, i.e., of the
intuition of our self and our inner state

–
the possibility of either simultaneity or
succession in the perception of objects.

Synthetic a priori

•
Quantity

unity, plurality, and totality

–
Also number

•
Quality

reality, negation and limitation

•
Modality

possibility
–

impossibility, existence
–

non
-
existence, necessity
–

contingency

•
Relation

inherence and subsistence, causality and
dependence, and community (reciprocity between
agent and patient)

Elements of formal ontology

•
From space
–

identity

•
From time
–

sequence

•
From quantity
–

representation structures, the
part/whole relationship, arithmetic

•
From quality
–

negation, unity, identity of
complex objects

•
From modality
–

formal logic

•
From community
–

entities and attributes,
dependence, causality, mutual exclusion and
complex objects


Formal vs material ontology

•
Formal

–
GOL

–
Ontoclean

–
BWW

•
Material

–
WordNet

•
Mixed

–
Cyc

–
SUMO

formal ontologies <
-

knowledge
representation

•
Programs = algorithms + data structures

•
KR systems follow the categories

•
Not so well as humans

•
Formal ontologies add richness

•
But are not qualitatively different

•
A tsunami is not a partial differential equation



So what?

•
Material ontologies fail due to semantic
heterogeneity

•
Formal ontologies are neutral wrt content

–
More limited aim

•
Provide rich abstract data types

•
Do not resolve semantic heterogeneity


Final thought

•
Same entity can be represented differently for
different purposes. Eg, aircraft might be

–
Part/whole to manufacturer (bill of materials)

–
Set/instance to airline (seats)

–
Process/accomplishments to production scheduling

–
Physical object/ qualities to inspection system

•
But one view might incorporate elements of all of
these


Final thought

•
Formal ontologies not content

–
Formal concepts should not be at top of
subsumption structures

•
A tsunami is not a partial differential equation

•
A bill of materials is not a part
-
whole system

–
But using schemas/ variable instantiation

–
Separate formal and material as facets