KR-2002 Panel/Debate

1
KR-2002 Panel/Debate

• Are Upper-Level Ontologies worth the effort?
• Chris Welty, IBM Research

2
What is Ontology?Smith Welty, 2001

• A discipline of Philosophy
• Meta-physics dates back to Artistotle
• Ontology dates back to 17th century
• The science of what is
• One universe - one ontology
• Borrowed by AI community
• McCarthy (1980) calls for a list of things that
  exist
• Evolution of meaning
• Now refers to domain modeling, conceptual
  modeling, knowledge engineering, etc.

3
What is an Ontology?Welty, Uschold, Gruninger,
Lehmann McGuinness, 1999
4
What is an Upper Level Ontology?

• The most general classifications of entities
• Domain independent
• High reusability, utility
• Common reference ontology
• The most basic building blocks (foundation) of
  ontologies, e.g.
• Properties Event/object, abstract/concrete,
• Relations Isa, part, instance,

5
Examples of Upper Level Ontologies

• Wordnet Upper Level Miller, Fellbaum
• Cyc Upper Level Lehmann, Klein
• IEEE SUO
• OntoClean Top Level Guarino, et al
• GOL Smith, et al

6
What an Upper Level Ontology is not

• Necessarily large
• Necessarily part of any system that uses it
• Necessarily goal-independent
• Necessarily language dependent
• Easy

7
Panelists

• Pro
• Nicola Guarino
• CNR, Italy
• Brandon Bennett
• U. Leeds, UK

• Con
• Alan Rector
• U. Manchester, UK
• Jim Hendler
• U. Maryland, USA

8
Panelist IntroductionNicola Guarino

• Background
• Ontology, Formal Ontology
• Recent Work
• OntoClean, restructuring WordNet
• Goal
• Bring together CS, Philosophy, and Linguistics
• Ontology
• Account of intended meaning of terms

9
Panelist IntroductionBrandon Bennett

• Background
• Spatial and temporal reasoning
• Recent Work
• Axiomatic systems for space/time, vagueness VUG
• Goal
• A unifying foundational ontology
• Ontology
• Rigorous definition of concepts

10
Panelist IntroductionAlan Rector

• Background
• Medical terminologies, description logics
• Recent Work
• GALEN, part/whole reasoning, drug info
• Goal
• Integration of clinical systems with decision
  support, indexing of evidence
• Ontology
• Implementation of the formal representation of
  concepts needed for a collection of information
  resources

11
Panelist IntroductionJim Hendler

• Background
• Scaleable KR Systems, Interoperability, Agents
• Recent Work
• DAML, DAMLOIL, W3C webont chair
• Goal
• A semantic web
• Ontology
• A web of nested terminologies

12
Are upper-level ontologies worth the effort?
13
Points Against

• No The
• Agreement impossible
• On the scale of the web, ridiculous
• Standardizing syntax OK, but not semantics
• Cognitive systems differ, even within people
• Different cog system, different ontology
• Consequences of decisions hard to predict, but
  important
• Middle level easier, more important
• Upper level development often gets bogged down
• Takes time away from more critical parts of a
  system
• No way to evaluate quality
• Easy to teach someone middle-level concepts
• Upper level often too opaque to explain
• Some domains require different levels of
  complexity depending on the needs

14
Points for

• Standardizing upper level is no different than
  standardizing syntax, its just the next step
• to establish consensus, there must be some common
  ground
• We dont know how to argue
• Need tools to support argumentation
• We must understand when we disagree

15
Other points

• What is upper level?
• Abstract, or simple?
• Where is the evidence to support upper level
  notions?
• Time easy or hard?
• How relevant is philosophical ontology?
• What is the value of legislation?
• Role of logic, category theory
• How do we learn how to communicate
• Upper ontology similar to language creation?