Semantic Mediation Tools

1
Semantic Mediation Tools

• Mala Mehrotra
• Pragati Synergetic Research, Inc.
• Cupertino, CA
• mm_at_pragati-inc.com

Extended Metadata Registry (XMDR) Project
Meeting Jan 25-28, 2005 UC Berkeley Faculty Club
2
Outline

• Motivation
• Pragati Tool Suite Overview
• MVP-CA Technology Core
• MVP-CA Results
• Knowledge Entry Aid (K)
• Quality Assurance Aid (Q)
• Mapping Merging Aid (M)
• OSRT Overview (ONR SBIR)
• COE Plug-In Overview (DOD/IHMC)
• IOD Overview (NASA SBIR)
• Conclusion

3
Ontologies Enable Interoperability Collaboration
An ontology is

• Set of concepts used in a particular domain of
  discourse
• Interrelationships between the concepts
• behavioral characteristics
• operational characteristics

• Ontologies provide a framework for expressing a
  common lingua to be utilized across systems
• Expressing semantics of concepts formally through
  ontologies enables automated reasoning, desired
  functionalities and collaboration across systems

4
Ontological Design Principles

• Ontological engineers try to optimize the
  ontological design
• Parsimonious design of concept classes
• Crispness in the distinctions across concepts
• Richness in the associations across concepts

5
Ontological Concerns

• Information overload is occurring in the creation
  of ontologies
• Every organization thinks their core ontology
  will be the Holy Grail for ontologies
• Reality 1 The notion of a canonical ontology
  is, at least at present, a myth
• Reality 2 We currently have to live with a
  cloud of candidate ontologies which model a
  real concept from different perspectives

Ontology Developers Dilemma How can I
effectively find and reuse concepts from that
cloud?
6
Pragatis Expertise

• Tools to support the analysis, quality-assurance
  and reuse of axiom-based ontologies
• Multi-ViewPoint Clustering Analysis (MVP-CA)Tool
• Ontology Search Reuse Tool (OSRT)
• COE (Collaborative Ontology Environment)
• Applied the tools to ontologies from a wide range
  of application domains, including
• Core ontologies (Cyc, KM Darpas RKF program)
• Biology
• Military
• Telemetry systems
• Diagnostic systems

7
Pragatis Solution

• Pragatis expertise and technology can provide
  key support in the areas of
• Development Knowledge entry aids
• Maintenance Quality Assurance aids
• Interoperability Mapping/Merging aids
• of ontologies

8
Vision

• Build semantic mediation tools to aid
• Building of rich ontologies through knowledge
  analysis and knowledge capture
• Provide semantics-directed search for
  concept-reuse
• Efficient maintenance of ontologies by making
  ontology aware quality-assurance tools
• Aid conflict-discovery
• Expose implicit design artifacts in ontologies
  through reverse engineering
• Meaningful interoperation/integration of
  ontologies by exposing potential mapping regions
  through similarity analysis

9
Pragati Tool Suite
Inter-Ontology Mappings
New Ontologies
Adapted Ontologies
OpenCyc
IOM
C2IEDM
IOD
COE Plugin
OSRT
ASRS
Core MVP-CA Clustering Technology
AIDS
ATIS
10
XMDR/OMG Pragati Tools Connection
Ontology A (OWL)
Ontology B (SCL)
Ontology C (CLIPS)
Ontology D (ER)
OWL Connector
SCL Connector
CLIPS Connector
ER Connector
Metadata Registry
Cross-Language Support (UML/DL/KIF-based?)
Pragati Tool Suite
B
D
A
C
11
Core technology MultiViewpoint Clustering
Analysis (MVP-CA)

12
Multi-ViewPoint Clustering Analysis

• Approach
• Cluster a knowledge base from multiple
  perspectives

• Clustering of knowledge bases into groups of
  semantically-related rules/axioms reveals
• Relationship of terms in the context of their
  usage
• Prototypical patterns of usage for the terms in
  the axioms
• Multiple ways of clustering (based on different
  objective criteria) aid in understanding and
  analyzing KBs from different perspectives

13
MVP-CA Architecture
KB/Ontology
Parsing Phase
Parser
Rule Filters
Pattern Filters
file.parsed
Clustering Tool
Cluster Generation Phase
Distance Metric
Pattern Filters
file.clustered
Other Tools
Analysis Modules
Web Services API
GUI
IOD
Conc. Extr.
VV
OSRT
Cluster Analysis Phase
Ranking
COE
Vicinity
Templating
Shaken
14
Knowledge Bases Analyzed

• OWL Ontologies (Wine, Time, FOAF, GeoRelations,
  etc.)
• Aviation-Related Ontology Reports
• Aviation Safety Reporting System (ASRS)
• Inflight Weather Encounters
• Automated Terminal Information Service (ATIS)
• Combined Reports for BOS, LAX, SFO,
• FAA Accident / Incident Data System (AIDS)
• 200 reports involving LAX
• IKB Micro Theories (Cycorp Inc) (CycL)
• CPoF Microtheories
• Spatial Axioms
• Molecular Biology Mt
• Chemistry Mt
• Biochemistry Mt
• Shaken and Kraken KBs (DARPAs RKF Program) (MELD
  Knowledge Machine(UT))
• transcription, translation and replication
  processes in molecular cell biology
• COA Critiquing Challenge Problem from RKF
• CMCP Crisis Management Challenge Problem
  (DARPAs HPKB Program) (KIF)
• problem solving for influencing national security
  and defense interests

15
Knowledge Bases Analyzed

• IMMACCS Integrated Marine Multi-Agent Command
  Control System (Marine ONR) (COOL)
• real-time decision-support system for
  situation-awareness
• ONAV Onboard NAVigation System (NASA Johnson)
  (CLIPS)
• navigates reentry of spacecraft into earths
  atmosphere
• XTE X-Ray Timing Explorer (NASA Goddard)
  (GENSAA)
• Health Safety Monitoring Rule base for various
  onboard subsystems
• UES Unexpected Events System (Interface and
  Control Systems) (SCL)
• Part of the NASAs Far Ultra-Violet Spectroscopic
  Explorer (FUSE) project
• SEA-ES Spacecraft Environmental Anomalies
  (Aerospace) (TIs Personal Consultant Plus)
• Detection of on-orbit satellite anomalies due to
  space environmental effects
• ESAA Expert System Advocate Advisor (DOT)
  (EXSYS)
• Advocates conditions for building an expert
  system
• PAMEX Pavement Maintenance Expert System (DOT)
  (EXSYS)
• Advocates repair actions for different types of
  pavement damages

16
Ontological Issues

• Level of Abstraction
• Concepts are too specialized
• May need intermediate concepts to capture the
  commonalities
• Example Ford Taurus, Toyota Camry, Honda
  Accord gt Automobiles
• Concept is too general
• May need to be specialized
• Example Move gt
• Move-Into, Move-To, Move-Out-Of,
  Move-Through

17
Ontological Issues (contd.)

• Presence of similar terms which may not share
  proximity in the ontology
• Lexically and semantically close terms
• Example Move Move-Into,
• Touches TouchesDirectly
• Prevent Prevents
• Lexically distant but semantically close terms
• Example Move-Into Enter
• Enter Penetrate
• Location Place
• Overloading of semantics in composite terms
• Example PortalOpen OpenPortal
• HoldTogether HoldInPlace
• Complementary terms
• Example Move-From Move-To

18
Ontological Issues (contd.)

• Context of usage
• Concept attributes
• Attributes defined for the concept based on its
  usage
• Example contract -gt physical change vs. legal
  document
• culture -gt societal issues vs. biological
  experiment
• Placement in the hierarchy
• Choice of axis of distinction when you have
  orthogonal characteristics
• Example organizing clothes line for a
  department store layout vs. for a manufacturer
  according to
• gender (mens, womens) vs. clothes type
  (pants, shirts)
• Recognize Capture Design Patterns in Concept
  Definitions

19
Key Contribution Areas
20
Pragatis Contribution Areas

• Knowledge Entry (K) Aid
• KE aid during the formation of the core and
  application ontologies
• Quality Assurance (Q)
• Analyzing the ontologies for errors
• Mapping/Merging (M)
• Providing support during mapping of application
  ontologies onto the core ontology

21
Knowledge Entry Aid

• Knowledge Discovery
• Discovering the relevant concepts to be entered
• Disambiguating concepts
• Context relevance
• Capturing design patterns in concept definitions

22
KE Aid Knowledge Discovery
23
Extracted Software Architecture(Spatial Slice
CycL)
Orientation
orientation of sorts
direction
geographicalDirection horizontalDirection  
24
Context discovery for a core concept Upper Cycs
Spatial Slice
Allow specialization of over-used concept terms
to reflect different contexts in which they are
utilized.

near
inAmong
covers-Hairlike
touches
groupMembers
groupMembership
onPhysical
objectFoundInLocation
positional
touchesDirectly
fluids
partonomic
in-Floating
surfaceParts
geographic
BodyOfWater
physicalParts
containsCavity
in-ImmersedFully
geographicalSubRegions
GeographicalRegion
in-ContGeneric
in-ImmersedGeneric

• The concept can be specialized based on
• the type of object to be found
• the location in which one is trying to find the
  object
• the precise type of find that defines success.

25
Axiom Cluster for ObjectFoundInLocation
(implies ( objectFoundInLocation ?OBJ
?LOC)(near ?LOC ?OBJ)) (implies(and
(touchesDirectly ?X ?Y)(objectFoundInLocation
?X ?LOC)) (objectFoundInLocation ?Y
?LOC)) (implies(and (touches ?X
?Y)(objectFoundInLocation ?X ?LOC)) (objectFo
undInLocation ?Y ?LOC)) (implies(and(on-Phys
ical ?X ?Y)(objectFoundInLocation ?Y
?LOC)) (objectFoundInLocation ?X
?LOC)) (implies(and (groupMembers ?C
?MEM)(objectFoundInLocation ?C
?LOC)) (objectFoundInLocation ?MEM
?LOC)) (implies (and (physicalParts ?X
?PART)(objectFoundInLocation ?X
?LOC)) (objectFoundInLocation ?PART ?LOC))
(implies (and (physicalParts ?LOC
?PART)(objectFoundInLocation ?X
?PART)) (objectFoundInLocation ?X
?LOC)) (implies (and (in-ContGeneric ?OBJ
?CONT)(objectFoundInLocation ?CONT
?REG)) (objectFoundInLocation ?OBJ
?REG)) (implies (in-ImmersedFully ?OBJ
?FLU) (objectFoundInLocation ?OBJ
?FLU)) (implies (and (isa ?FLUID
Place)(in-ImmersedGeneric ?OBJECT
?FLUID)) (objectFoundInLocation ?OBJECT
?FLUID)) (implies (and(objectFoundInLocation
?PER ?LOC)(covers-Hairlike ?STUFF
?LOC)) (in-Among ?PER ?STUFF)) (implies(and
(in-Floating ?OB ?LIQ)(surfaceParts ?LIQ
?SURF)) (objectFoundInLocation ?OB
?SURF)) (implies (and (isa ?WATER
BodyOfWater)(in-Floating ?OBJ
?WATER)) (objectFoundInLocation ?OBJ
?WATER)) (implies (and (in-ContGeneric ?OBJ
?CONT)(containsCavity ?CONT ?CAV)) (objectFou
ndInLocation ?OBJ ?CAV)) (implies(geographical
SubRegions ?REG ?PLACE) (objectFoundInLocation
?PLACE ?REG)) (implies (and (isa ?Y
GeographicalRegion)(on-Physical ?X
?Y)) (objectFoundInLocation ?X ?Y))
26
Discover Choices for Ont. Hierarchy Placement for
ConceptCycs BioChemistry Mt.

• A nucleotide molecule can be represented by
• holding the sugars constant at first level and
  varying the base (left figure) or
• holding the base constant at first level and
  varying the sugar (right figure)
• The left representation good for chain type of
  reasoning for the molecule that is at the
  nucleotide level.
• The right representation good for the matching
  base pair type of level of reasoning.
• Clustering brought to attention both these
  representations.

Sugar-dependent representation
Base-dependent representation
27
Discover similar knowledge entry patterns
Templates
28
Generalized Concept Formation (Cyc Spatial Slice)

• Abstract Spatial Relationships Cluster
• Axioms are expressing uniqueness of ConvexHullFn
  and InteriorFn
• Concept has wider applicability hence
  generalize through templatization

(implies (and (termOfUnit ?CONVEXHULLFN
(ConvexHullFn ?OBJECT)) (termOfUnit
?CONVEXHULLFN-1 (ConvexHullFn
?CONVEXHULLFN))) (equals ?CONVEXHULLFN
?CONVEXHULLFN-1)) (implies (and (termOfU
nit ?INTERIORFN (InteriorFn ?INTERIORFN-1)) (
termOfUnit ?INTERIORFN-1 (InteriorFn
?ANYOBJECT))) (equals ?INTERIORFN
?INTERIORFN-1))
Template (implies (and (ltUniqueFngt
?ltUNIQUEFNgt) (termOfUnit ?ltUNIQUEFN-TERM-1gt
(?ltUNIQUEFNgt ?OBJECT)) (termOfUnit
?ltUNIQUEFN-TERM-2gt (?ltUNIQUEFNgt
?ltUNIQUEFN-RESULT-1gt))) (equals
?ltUNIQUEFN-TERM-1gt ?ltUNIQUEFN-TERM-2gt))
29
Generalizing a COA Critiquing Concept
Armor-Brigade-Attacking-Mechanical-Battalion
Armor-Brigade-Attacking-Armored-Battalion
Big-Unit-Attacking-Small-Unit
30
Intermediate Concepts by Factoring IMMACCS
TrackPosition
LethalWeapon
WeaponSelection
Platform
Dimension
Munitions
Agent
Entity
CallFor Fire
Position
Rotary Wing
Structure
Cluster conflict due to blocking rotary
wing Rules Structure_Trajectory_Weapon
Structure_Trajectory_Entity
Structure_Trajectory_Platform
Cluster conflict due to blocking building Rules
RotaryWing_Trajectory_Weapon
RotaryWing_Trajectory_Entity
RotaryWing_Trajectory_Platform

• Objects shown in green are used by the two sets
  of rules performing conflict resolution these
  can be factored into an intermediate object.
• The new intermediate object would be utilized by
  both sets of rules and other rules addressing
  similar conflict situations.

31
Axiom Interrelationships Capture (Shaken-Core)
(every Color-Value has (color-of ((must-be-a
Tangible-Entity))) (value ((possible-values
(the instances of (the
categorical-constant-class of color)))))
(same-as ((must-be-a Color-Value)))) (every
Consistency-Value has (consistency-of
((must-be-a Tangible-Entity))) (value
((possible-values (the instances of
(the categorical-constant-class of
consistency))))) (same-as ((must-be-a
Consistency-Value)))) ... (every ltxgt-Value has
(ltxgt-of ((must-be-a Tangible-Entity))) (value
((possible-values (the instances of
(the categorical-constant-class of
ltxgt))))) (same-as ((must-be-a ltxgt-Value)))) X
gets instantiated with values such as direction,
manner, smell, sex, trait, etc
(every Brightness-Value has (brightness-of
((must-be-a Tangible-Entity))) (less-than
((must-be-a Brightness-Value))) (greater-than
((must-be-a Brightness-Value))) (same-as
((must-be-a Brightness-Value)))) (every
Capacity-Value has (capacity-of ((must-be-a
Tangible-Entity))) (less-than ((must-be-a
Capacity-Value))) (greater-than ((must-be-a
Capacity-Value))) (same-as ((must-be-a
Capacity-Value)))) ... (every ltxgt-Value has
(ltxgt-of ((must-be-a Tangible-Entity)))
(less-than ((must-be-a ltxgt-Value)))
(greater-than ((must-be-a ltxgt-Value)))
(same-as ((must-be-a ltxgt-Value)))) X gets
instantiated with values such as age, area,
density, depth, etc
32
Lexically Removed Semantically Close Terms
Similar Slots used in KM Core Spatial-Entity,
Place Move

• Clustering exposed a long list of shared slots
  across the three concepts
• is-near, abuts, is-above, is-below, etc (18 slots
  common to 3 axioms)
• Place has additional slots like is-between and
  is-beside
• Move makes use of slots in both Spatial-Entity
  and Place
• Place is a subclass of Spatial-Entity
• Potential maintenance problem each of the slots
  must be repeatedly dealt every time
  Spatial-Entity is specialized (Place) or
  manipulated (Move)

Possible Solution Parameterize the repetitive
slot definitions through templates
33
Parameterization of slot-propagation
(every Spatial-Entity has (location ((must-be-a
Place))) (is-near ((must-be-a
Spatial-Entity))) (abuts ((must-be-a
Spatial-Entity) (excluded-values
(Self)))) (is-outside ((must-be-a
Spatial-Entity))) (does-not-enclose ((must-be-a
Spatial-Entity))) (is-inside ((must-be-a
Spatial-Entity))) )
(every Place has(location ((exactly 0
Place)))(is-near ((forall (the location-of of
Self) (the is-near of It))))(abuts ((forall (the
location-of of Self) (the abuts of
It)))) (is-beside ((forall (the location-of of
Self) (the is-beside of It))))(is-between
((forall (the location-of of Self) (the
is-between of It))))...)
34
Parameterization of slot-propagation
(every Move has... (destination ((must-be-a
Spatial-Entity))) (add-list ((if
(has-value (the destination of Self))
then (forall (the
object of Self) (set
(triple It is-near (the
is-near of (the destination of Self)))
(triple It abuts (the abuts of
(the destination of Self)))
(triple It is-beside (the is-beside of (the
destination of Self))) (triple
It is-between (the is-between of (the destination
of Self))) ... )))))(del-list ((forall (the
object of Self) (set
(triple It location (the location
of It)) (forall2 (the
is-near of It) (if (not
((the is-near of (the destination of Self))
includes It2)) then
(triple It is-near It2))) (forall2
(the abuts of It) (if
(not ((the abuts of (the destination of Self))
includes It2)) then
(triple It abuts It2))) (forall2 (the
is-beside of It) (if (not
((the is-beside of (the destination of Self))
includes It2)) then
(triple It is-beside It2))) (forall2
(the is-between of It)
(if (not ((the is-between of (the destination of
Self)) includes It2))
then (triple It is-between It2))) ... )))))
35
Higher-level axiom forSpatial-Entity, Place
Move
Template 1 (ltpropertygt ((forall (the location-of
of Self) (the ltpropertygt of It)))) Template 2
(triple It ltpropertygt (the ltpropertygt of
(the destination of Self))) Template 3 (forall2
(the ltpropertygt of It) (if
(not ((the ltpropertygt of (the destination of
Self)) includes
It2)) then (triple It
ltpropertygt It2)))
Higher Level Axiom If Spatial-Entity has slot
with ltpropertygt then Place should have clause
ltTemplate 1gt and Move should have clause
ltTemplate 2gt in add-list and ltTemplate
3gt in del-list
36
Reification of Slot Propagation Divide
Duplicate

• Divide and Duplicate belong to two different
  branches in the ontology as shown
• Divide and Duplicate brought together due to
  access of similar slots
• material, age, animacy, area, breakability, etc.
• The slots of Duplicate propagate as is to the sub
  object straight propagation
• Whereas for Divide
• some slots utilize straight propagation as above
• animacy, breakability, temperature, taste,
  texture
• other slots have decreasing propagation
• age, area, depth, height, length, etc.
• Mode of attribute propagation can be reused in
  different situations
• Suggested solution Reify the various propagation
  modes
• ltxgt propagates to ltclass-objectgt using ltmode-xgt
  propagation

37
Propagation of Discrete vs. Continuous Values
(every Brightness-Value has (brightness-of
((must-be-a Tangible-Entity))) (less-than
((must-be-a Brightness-Value))) (greater-than
((must-be-a Brightness-Value))) (same-as
((must-be-a Brightness-Value)))) (every
Capacity-Value has (capacity-of ((must-be-a
Tangible-Entity))) (less-than ((must-be-a
Capacity-Value))) (greater-than ((must-be-a
Capacity-Value))) (same-as ((must-be-a
Capacity-Value)))) ... (every ltxgt-Value has
(ltxgt-of ((must-be-a Tangible-Entity)))
(less-than ((must-be-a ltxgt-Value)))
(greater-than ((must-be-a ltxgt-Value)))
(same-as ((must-be-a ltxgt-Value)))) X gets
instantiated with values such as age, area,
density, depth, etc
(every Color-Value has (color-of ((must-be-a
Tangible-Entity))) (value ((possible-values
(the instances of (the
categorical-constant-class of color)))))
(same-as ((must-be-a Color-Value)))) (every
Consistency-Value has (consistency-of
((must-be-a Tangible-Entity))) (value
((possible-values (the instances of
(the categorical-constant-class of
consistency))))) (same-as ((must-be-a
Consistency-Value)))) ... (every ltxgt-Value has
(ltxgt-of ((must-be-a Tangible-Entity))) (value
((possible-values (the instances of
(the categorical-constant-class of
ltxgt))))) (same-as ((must-be-a ltxgt-Value)))) X
gets instantiated with values such as direction,
manner, smell, sex, trait, etc
38
Inverse Classes

• Clichés is a promising implementation mechanism
• Different types of inverse classes
• Change In Degree Increase Decrease
  (greater-than and less-than)
• Reflexive Move-Together Move-Apart
• (reversal of origin and destination slots)
• Exit Enter
• (agent and object slots identical)
• Near-inverse classes
• Come-Together/Disperse
• In addition to reversal of origin and destination
  slots reference
• Different sub-event classes
• Come-Together has Go-To sub-event
• Disperse has Leave sub-event

39
Pragatis Contribution Areas

• Knowledge Entry (K) Aid
• KE aid during the formation of the core and
  application ontologies
• Quality Assurance (Q)
• Analyzing the ontologies for errors
• Mapping/Merging (M)
• Providing support during mapping of application
  ontologies onto the core ontology

40
Ontology Hierarchy Placement Error (KM-Core)
(every Aperture has (in-event ((a
Move-Through))) (played-by ((a
Spatial-Entity with
(path-of ((the in-event of Self))))))) (every
Passage has (in-event ((a
Move-Through))) (played-by ((a
Spatial-Entity with
(path-of ((the in-event of Self))))))) (every
Conduit has (in-event ((a Move)))
(played-by ((a Entity with
(path-of ((the in-event of Self)))))))
(every Egress has (in-event ((must-be-a
Move-Out-Of))) (played-by ((a
Spatial-Entity with
(path-of ((the in-event of Self))))))
(is-between ((args (the is-inside of Self)
(the is-outside of
Self))))) (every Entrance has (in-event
((must-be-a Move-Into))) (played-by ((a
Spatial-Entity with
(path-of ((the in-event of Self))))))
(is-between ((args (the is-inside of Self)
(the is-outside of Self)))))

• This cluster defines concepts that are
• subclasses of Role
• related by different types of Move events
• Conduit specifies an Entity for the played-by
  slot instead of Spatial-Entity.
• Is this intentional?

41
Inconsistent Axioms (UES from NASAs FUSE
Project)
rule modal_im_det2auxpwrst -- rule 25 subsystem
modal_i_det_t27hsk category rtmm if change
(i_det2auxpwrst) then if i_det2auxpwrst
0 then msg "modal tlm
i_det2auxpwrst now in mode 1" else
if i_det2auxpwrst 1 then
msg "modal tlm i_det2auxpwrst now in mode
2" . end modal_im_det2auxpwrst
rule modal_im_det2auxpwrst -- rule 26 subsystem
modal_i_det_t27hsk category rtmm if change
(i_det2auxpwrst) then if i_det2auxpwrst
1 then msg "modal tlm
i_det2auxpwrst now in mode 1" else
if i_det2auxpwrst 0 then
msg "modal tlm i_det2auxpwrst now in mode
2" ... end modal_im_det2auxpwrst
42
Typographical Errors Inter-cluster Relationships
Capture (XTE NASA)
Rule Rule Description 23
sa_status_check 27 xpndr_status_check 29
gsace_status_check 33 tam_status_check 35
tam_status_check 25 sds_status_check 31
rwa_status_check
Rule Rule Description 24 sa_limit_check
28 xpndr_limit_check 30 gsace_limit_check
34 tam_limit_check 36 pca_limit_check 26
sds_limit_check 32 rwa_limit_check
Rule 35 (defrule tam_status_check
"" (LimitStatus PAPCU1TMP2TXTE_DECOM
?x1) ... ?o1 lt- (Inferred PCA-Temp-Status
?cur_stat) (Inferred valid-telemetry valid) gt
(if (neq ?cur_stat ?new_stat)
then . then (SendMessage "MessageWindow"
Status (str-cat "PCA Temperatures changed
from " ?cur_stat " to " ?new_stat)) else
(SendMessage "MessageWindow" Warning (str-cat
"PCA Temperatures changed from " ?cur_stat " to "
?new_stat)) ...
43
Eliminate Logical Redundancy (Cyc Spatial-Core)
Category 1 Check potentially redundant axioms
through transitivity.

• Immersion Cluster
• surroundsHorizontally is being deduced in two
  different ways from in-ImmersedPartly
• in-ImmersedGeneric is redundant for establishing
  this relationship
• (implies(in-ImmersedPartly ?OBJ
  ?FLUID)(in-ImmersedGeneric ?OBJ ?FLUID))
• (implies(in-ImmersedGeneric ?OBJ
  ?FLUID)(surroundsHorizontally ?FLUID ?OBJ))
• (implies(in-ImmersedPartly ?OBJ
  ?FLUID)(surroundsHorizontally ?FLUID ?OBJ))

44
Eliminate Declarative Redundancy (Cyc
Spatial-Core)
Category 2 Check redundancy because of
pre-existing ontological declarations using Cycs
powerful predicates

• Immersion Cluster
• touches is already a genlPreds for
  in-ImmersedGeneric
• (implies(in-ImmersedGeneric ?OBJ ?FLUID)
  (touches ?FLUID ?OBJ))
• in-ImmersedFully is already a genlPreds for
  suspendedIn
• (implies(suspendedIn ?OBJ
  ?FLU)(in-ImmersedFully ?OBJ ?FLU))
• in-ImmersedGeneric is already a genlPreds for
  in-ImmersedFully
• (implies(in-ImmersedFully ?OBJ
  ?FLUID)(in-ImmersedGeneric ?OBJ ?FLUID))

• Spatial Properties Cluster
• BorderbetweenFn is already a declared commutative
  function
• (implies(and
• (termOfUnit ?BORDERBETWEENFN
  (BorderBetweenFn ?REG2 ?REG1))
• (termOfUnit ?BORDERBETWEENFN-1
  (BorderBetweenFn ?REG1 ?REG2)))
• (equals ?BORDERBETWEENFN ?BORDERBETWEENFN-1))

45
Eliminate Multiple Versions of Concept (Single
Author Shaken COA Critiquing)

• It is not clear which version is intended to be
  retained. Need to weed out deadwood critiquing
  pattern.
• There is more information carried in the
  PatternBridgesRivers axiom. Bridge is-on the
  MainAttackAxis in PatternBridges-Rivers.
• PatternBridges-Rivers-2 appears to be a bit less
  restrictive in that it doesnt require the bridge
  being seized to be on the main attack path.

46
Eliminate Multiple Versions of Concept (Multiple
Authors Shaken COA Critiquing)
47
Semantic incompleteness (Cyc Spatial-Core)

• Borders Cluster
• spatiallyIntersects seems to be incompletely
  specified w.r.t. ?D
• Either
• Two other sets of rules with ?D in consequent are
  needed, or
• Last or condition needs to have ?D instead of
  ?C
• Authors intentions are not clear

(implies (and (formsBorderBetween
?BORDER ?A ?B) (formsBorderBetween ?BORDER ?C
?D)) (or (spatiallyIntersects ?A
?C) (spatiallyIntersects ?B ?C)))
48
Pragatis Contribution Areas

• Knowledge Entry (K) Aid
• KE aid during the formation of the core and
  application ontologies
• Quality Assurance (Q)
• Analyzing the ontologies for errors
• Mapping/Merging (M)
• Providing support during mapping of application
  ontologies onto the core ontology

49
Mapping Aids

• Support for mapping of terms across ontologies
  can be provided by highlighting terms that are
  used in similar contexts
• Such similarities can be extracted using the same
  technology as used for template formulation and
  redundancy detection when applied across
  different ontologies

50
Similar Concepts Lexically Close (Cycs
Spatial-Core)
(implies (and (touchesDirectly ?X
?Y)(objectFoundInLocation ?X ?LOC)) (objectFo
undInLocation ?Y ?LOC)) (implies (and
(touches ?X ?Y)(objectFoundInLocation ?X
?LOC)) (objectFoundInLocation ?Y ?LOC))
touchesDirectly and touches are essentially same
concepts in the context of objectFoundInLocation
and can be mapped to each other
51
Ontology Search Reuse Tool (OSRT)

52
OSRT Vision

• A tool that enables builders of knowledge-based
  systems to identify and reuse relevant portions
  of existing systems, thereby
• Reducing development time
• Amortizing development costs
• Enhancing quality of developed system
• overall increase in return on investment (ROI)

53
Ontology Search and Reuse Tool
CycL
ER
CLIPS
OWL
OSRT
54
Ontology Developers Dilemma

• Where is the concept?
• Searching for the relevant concept
• How is it used?
• Concept perspectives based on context of usage
• How to adapt it?
• Concept transformation and merging

55
Queries

• Semantically Rich Queries
• Concept Name
• Attribute Name
• Generalization Structure
• Association Relationship
• Vicinity Concept
• Repertoire of String Matching Algorithms
• Component Vector Overlap
• Substring Matching
• Query Plug-In support
• To allow new types of queries to be easily
  integrated into the framework

56
Concept Usage Views

• Cognitive Aids for Concept Selection
• Definitions View
• Displays the focus concept as declared in the
  ontological hierarchy through Embarcadero
  Describes XMI export
• Vicinity Concepts View
• Displays the vicinity concepts concepts that
  co-occur with the focus concept
• Rules Usage View
• Displays the cluster of rules where the focus
  concept has localized
• Templates View
• Displays the templates associated with the
  cluster of rules

57
Query Based on Concept Name
58
Vicinity Concepts View
59
Adaptation Support

• Concept Adaptation
• Copy and paste into target ontology
• Edit concept attributes and relationships
• Merge with concepts in the target ontology
• Rules Adaptation
• Display MVP-CA generated templates in OSRT

60
Concept Adaptation in Target Ontology
61
Concepts Merging in Target Ontology
62
Collaborative Ontology Environment (COE)-Plug-In

63
COE Environment (IHMC)
A Collaborative Environment for Viewing,
Searching and Developing OWL Ontologies using
CMAPs

• Ontology Viewer
• Transforms ontologies written in OWL-family into
  natural CMAPs
• Suppresses mundane/obvious information
• Determines graph layout to show CMAPs
• Ontology Search
• Searches ontologies locally on web
• Mechanisms to book mark ontologies
• Support for searching for concepts in these
  ontologies
• Ontology Development
• Drag drop support for incorporating concepts in
  existing ontologies
• CMAP tools for graphical editing of concepts
• Transformation tools from CMAP to
  XML/RDF/DAML/OWL format
• Real-time collaboration aids for geographically
  distributed groups

64
Constructing ontologies as concept maps

• Exporting concept maps to OWL format.
• Concept map conventions for defining
  restrictions.
• atleast, atMost, must be,
• Templates aid in forming most repetitive,
  complicated restrictions that can be exported to
  OWL.

65
COE Issues

• User composing an ontology requires a concept.
• Invent one or re-use one?
• No semantic underpinnings for searching of
  concepts
• No cross ontology awareness
• Information overload when searching for relevant
  concepts
• No indication of relevance ranking in the
  retrieved ontologies
• COE tool will interface with Pragatis MVP-CA
  system to
• allow users to discover relevant concepts,
• across multiple ontologies and
• analyze them for reusability in context

66
COE/MVP-CA Integrated Interface
67
Current Status of Integration

• Can pull in clusters across multiple ontologies
  for a query term
• Clusters are currently being displayed from a
  pre-built repository
• 17 OWL ontologies clustered automatically (Wine,
  Ecoinformatics, Geo-related, Space
  Time-related, etc.)
• Can display stability-based information for a
  query term from COE
• Relationships across terms not immediately
  obvious for a given cluster
• Information overload for certain query terms,
  underload for others

68
Iterative Ontology Development Tool (IOD)

69
IOD Problem Statement

• Data and knowledge repositories contain
• Large amounts of unstructured but
• Stylized natural language text
• Simple text-based search techniques successful in
  retrieving somewhat relevant documents to a human
  analyst's needs,
• Information contained in those documents is
  opaque with respect to
• query
• manipulation
• reasoning tools
• semantic content of the text

Proposed Solution
Extract semantic content from the text and
capture it in an ontology
70
Approach

• Analyze the data set to generate sub-sets with
  related concepts/similar concepts
• Generate a regular expression to capture the
  similarity pattern
• Map the regular expression to an ontology
  fragment consisting of concepts from existing
  ontologies along with new concepts
• Use the extraction binding (regular expression
  and the mappings) to extract new instances from
  the data set

MVP-CA
IOD
IOD and Protégé
IOD and Protégé
71
Data-Flow between IOD, Protégé, Semantic Web
72
Mock-up of IOD Interface
73
Enable the Protégé/OWL Query Model

• Use a candidate OWL reasoning system such as JTP
  or RACER
• Query What is the mean duration of reported
  turbulence events?
• Answer a mean lower bound of 4 seconds, and a
  mean upper bound of 7.5 seconds.

74
Pragati Tool Suite
Inter-Ontology Mappings
New Ontologies
Adapted Ontologies
IOM
C2IEDM
IOD
CODE Plugin
OSRT
ASRS
Core MVP-CA Clustering Technology
AIDS
ATIS
75
Uniqueness of Overall Approach

• Allows subtle, semantically-oriented analysis of
  ontologies
• Pattern-based approach for clustering
• discovers pattern-conforming/non-conforming
  regions in KB
• Clustering in similarity space (instead of
  feature space)
• Reveals higher-level information on relationships
  across concepts
• Clustering axioms is based on usage of axioms
  (independent of the declared ontology)
• Reveals information of tacit nature not captured
  in the ontology
• Domain and representation-independent
• Allows flexibility in deploying technology to any
  semi-structured information system

76
Benefits

• Cost-Effective Solution for Building and
  Organizing Ontologies KBs
• Less time needed
• Less personnel needed
• Effective reuse of legacy systems
• Quality Solution enabling high-end analysis for
• Development
• Maintenance
• Interoperability
• Adaptive Solution to Changing Demands
• In time as ontologies evolve across applications
• In perspective for different types of users