COSC 4117 Artificial Intelligence

Document Sample
COSC 4117 Artificial Intelligence Powered By Docstoc
					  General Knowledge

Ontological background for everything
 Ontology of everyday knowledge
Mental events
   need to include representations of KB’s of
    other agents to reason about their plans
    and actions
     Wants (AgentSmith, Dead(Neo))
   need to ‘contain’ inconsistent knowledge
    to avoid interaction
     Believes(Gambler1, Faster(HorseA,HorseB))
     Believes(Gambler2,
      ~Faster(HorseA,HorseB))


                D Goforth - COSC 4117, fall 2003   2
 Ontology of everyday knowledge
Mental events
   PROBLEM:
     Wants (AgentSmith, Dead(Neo))
   Is Dead(Neo) a predicate or a term?




                D Goforth - COSC 4117, fall 2003   3
  Ontology of everyday knowledge
Mental events

      KB
              Believes(AgentX,“Rich(Paul)”)

                                    assumed KB              to reason about
                                    of AgentX as            what AgentX
~Rich(Paul)                            objects              believes,
                                                            run AgentX’s
                                “Rich(Paul)”                ‘interpreter’ with
                                                            postulated
                                                            reasoning powers



                         D Goforth - COSC 4117, fall 2003                   4
Ontology of everyday knowledge
 Time
   non-monotonic change in KB
   Frame problem – inferring what changes
    and what does not
   Actions as objects
   Reasoning about events, intervals
    fluent calculus



             D Goforth - COSC 4117, fall 2003   5
Ontology of everyday knowledge
 Default reasoning (missing information)
(more on defaults later)




              D Goforth - COSC 4117, fall 2003   6
Models of general knowledge
1. SUMO (Suggested Upper Merged
   Ontology)
   Alan Pease, IEEE standard
   Minimal – basis for adding domains
2. Cyc (“Sike”)
   Douglas Lenat, Cycorp
   Huge KB of common knowledge



              D Goforth - COSC 4117, fall 2003   7
 1. SUMO
 (Suggested Upper Merged Ontology)

 Written in FOL
 Approx 1000 concepts in ontology
 Useful for basis of ‘expert’ projects
  which do not need ‘common sense’
  knowledge
 Open source



                D Goforth - COSC 4117, fall 2003   8
  SUMO
  Base ontology – top-level ontology

                                 Entity

                      Physical            Abstract

                                          Quantity
               Object     Process          Attribute
                                             Relation
                                                Proposition
SelfConnectedObject     Region                     SetOrClass

Complete SUMO Ontology (PDF)
  SUMO
  Example sub-ontology Units of Measure

      PhysicalQuantity                                ConstantQuantity


      UnitOfMeasure                                   AngleMeasure

SystemeInternationalUnitOfMeasure                     PlaneAngleMeasure




                   D Goforth - COSC 4117, fall 2003                  10
 SUMO
 Equivalent to 2nd order power by treating
  functions, predicates, logical operators
    as objects, also (not real examples)
  F(x)         (apply F x)                         (function)
  P(x,y)       (holds P x y)                       (predicate)
  R(x)         (instance x R)                      (“)
  ( A B)      (infer AND A B)                     (logical)



                 D Goforth - COSC 4117, fall 2003            11
  SUMO example of logical - inverse
(instance inverse BinaryPredicate)
(instance inverse IrreflexiveRelation)
(instance inverse IntransitiveRelation)
(instance inverse SymmetricRelation)
(domain inverse 1 BinaryRelation)
(domain inverse 2 BinaryRelation)
(=>
    (inverse ?REL1 ?REL2)
    (forall (?INST1 ?INST2)
        (<=>
            (holds ?REL1 ?INST1 ?INST2)
            (holds ?REL2 ?INST2 ?INST1))))
EXAMPLE:
(inverse greaterThan lessThan)
                    D Goforth - COSC 4117, fall 2003   12
(subclass AnimacyAttribute BiologicalAttribute)
(exhaustiveAttribute AnimacyAttribute Living Dead)
(documentation AnimacyAttribute "&%Attributes that indicate whether an
&%Organism is alive or not.")
(instance Living AnimacyAttribute)
(documentation Living "This &%Attribute applies to &%Organisms that are alive.")
(=>
  (and
     (instance ?ORGANISM Organism)
     (agent ?PROCESS ?ORGANISM))
  (holdsDuring (WhenFn ?PROCESS) (attribute ?ORGANISM Living)))
(instance Dead AnimacyAttribute)
(subAttribute Dead Unconscious)
(contraryAttribute Dead Living)
(documentation Dead "This &%Attribute applies to &%Organisms that are
not alive.")
(=>
  (instance ?ORG Organism)
  (exists (?ATTR)
     (and
        (instance ?ATTR AnimacyAttribute)
        (attribute ?ORG ?ATTR))))
2. Cyc
   enCYClopedia
   Douglas Lenat
   Cycorp
   1984->
   general knowledge
      and common-sense reasoning


             D Goforth - COSC 4117, fall 2003   14
Cyc (from cyc.com)
 Ontology – 100,000’s of terms
 Millions of assertions   Description of ontology
   “Water is wet”               on cyc website

   “Everyone has a mother”
   “When you let go of things they usually
    fall.”
 Open version available – opencyc.com


                D Goforth - COSC 4117, fall 2003   15

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:9/29/2012
language:English
pages:15