A Collaborative Approach to Ontology Design

Document Sample
A Collaborative Approach to Ontology Design Powered By Docstoc
					  Communications of the ACM
February 2002 - special Ontologies issue
   A Collaborative Approach to
        Ontology Design

C.W.Holsapple, K.D.Joshi
    Approaches to ontology design
•   (Approach - Basis for Design)
•   Inspiration - individual viewpoint
•   Induction - specific case
•   Deduction - general principles
•   Synthesis - set of existing ontologies
•   Collaboration - multiple individuals’
    viewpoints (+ initial ontology - anchor)
    The collaborative approach
• Preparation
  – define design criteria
  – determine boundary conditions
  – determine evaluation standards
• Anchoring
  – specify initial ontology that will seed the
    collaborative effort
The collaborative approach (cont)
• Iterative Improvement - Delphi technique
  – identify diverse panel of participants
  – elicit their (independent) comments on the
    ontology
  – revise ontology addressing feedback
  – iterate until consensus
• Application
  – demonstrate use of ontology
 Predicting How Ontologies for
 the Semantic Web Will Evolve
• categories of ontology applications:
  –   neutral authoring
  –   ontology as specification
  –   common access to information
  –   ontology-based search
• 1st phase in the evolution of the semantic
  Web - development of decentralized,
  adaptive ontologies for software
  specification
  Making Ontologies Work for
 Resolving Redundancies Across
           Documents
• natural language processing stuff (Xerox)
Evaluating Ontological Decisions
      with ONTOCLEAN

N.Guarino, C.Welty
• OntoClean - methodology how to evaluate
  ontological decisions
• based on metaproperties (properties
  attached to classes, relations and properties
  that make up an ontology)
• essential property of an entity - must hold
  for it
  – property “being hard” is essential for hammers,
    but not for sponges
• special form of essence: rigid property -
  essential to all its instances
  – “being a person” - rigid
  – “being hard” - not rigid
• anti-rigid - never essential
  – “being a student”
• semi-rigid - all other properties
Student   anti-rigid




Person     rigid
• identity - enables recognizing individual
  entities as being the same
• unity - enables recognizing all parts that
  form an entity
 e.g., one hour    Time
                  duration

                             fails identity
                             analysis

e.g., 1:00-2:00    Time
today             interval
                                               Time
                                              duration




                                               Time
                                              interval
        Subsumption vs Constitution

                       not a whole,
                       not recognizable as an entity

              Water                                    Water

fails unity
analysis

              Ocean                                    Ocean
                      represents
                      whole objects
Subsumption vs Instantiation



Mammal       Species



          no, because different identity criteria
          it is instantiation
Human

           Chris
         Meta….



Mammal    Rigid class



         no, because different identity criteria

Human
           Chris
Subsumption vs Part/Whole


  Car                             Car

         no, because different
         essential properties

Engine                           Engine
Disjunction                              Polysemy


    Car part                                  Book

               no, because of                        no, because of
               rigidity criteria                     identity analysis
                               ...
                                      Bound            Abstract
Engine         Wheel
                                     volume             entity

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:14
posted:2/23/2012
language:
pages:18