OMEN: A Probabilistic Ontology Mapping Tool
description
Transcript of OMEN: A Probabilistic Ontology Mapping Tool
![Page 1: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/1.jpg)
OMEN: A Probabilistic OMEN: A Probabilistic Ontology Mapping ToolOntology Mapping Tool
Mitra et al.Mitra et al.
![Page 2: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/2.jpg)
The ProblemThe Problem
We need to map databases or We need to map databases or ontologiesontologies
Mapping of two different ontologies
![Page 3: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/3.jpg)
The ProblemThe Problem
Mapping is difficultMapping is difficult
Most mapping tools are impreciseMost mapping tools are imprecise
Even experts could be uncertainEven experts could be uncertain
We deal with probabilistic mappingsWe deal with probabilistic mappings
![Page 4: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/4.jpg)
The SolutionThe Solution
Infer mappings based on previous onesInfer mappings based on previous ones
We use Bayesian Nets for inferenceWe use Bayesian Nets for inference
We use other tools for initial We use other tools for initial
distributionsdistributions
Preliminary results are encouragingPreliminary results are encouraging
![Page 5: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/5.jpg)
Basic ConceptsBasic Concepts
Bayesian network:Bayesian network:
Probabilistic graphical model that Probabilistic graphical model that represents Random variablesrepresents Random variables
Evidence nodes: The value is givenEvidence nodes: The value is given
T
![Page 6: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/6.jpg)
Bayesian NetworkBayesian Network
Conditional Probability tables (CPT)Conditional Probability tables (CPT)
![Page 7: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/7.jpg)
Bayesian Nets in our Bayesian Nets in our approachapproach
How do we build the Bayesian NetHow do we build the Bayesian Net Nodes are property or class matchesNodes are property or class matches Classes are conceptsClasses are concepts Properties are attributes of classesProperties are attributes of classes
m(C1,C1’)C1 C1’Ontology 1 Ontology 2
![Page 8: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/8.jpg)
Building Bayesian NetsBuilding Bayesian Nets
![Page 9: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/9.jpg)
Our Bayesian NetsOur Bayesian Nets
All combinations of nodes is too All combinations of nodes is too manymany
We generate only “useful” nodes We generate only “useful” nodes The cutoff is k from evidence nodesThe cutoff is k from evidence nodes Up to 10 parents per nodeUp to 10 parents per node Cycles are avoided (confidence ~.5)Cycles are avoided (confidence ~.5)
![Page 10: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/10.jpg)
Our Bayesian NetsOur Bayesian Nets
We need evidence nodes and CPTsWe need evidence nodes and CPTs
Evidence nodes come from Evidence nodes come from
initializationinitialization
CPTs come from Meta-rulesCPTs come from Meta-rules
![Page 11: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/11.jpg)
Meta-rulesMeta-rules Describes how other rules should be usedDescribes how other rules should be used Basic Meta-ruleBasic Meta-rule
m(C1,C1’)C1 C1’
m(C2,C2’)C2 C2’
q q’
P1=x
P2=x+c
![Page 12: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/12.jpg)
Other Meta-rulesOther Meta-rules
Range: Restriction of property valuesRange: Restriction of property values
Mappings between properties and Mappings between properties and
ranges of propertiesranges of properties
Single rangeSingle range
SpecializationSpecialization
![Page 13: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/13.jpg)
Other Meta-rulesOther Meta-rules
Mappings between super classesMappings between super classesChildren matching depends on parents Children matching depends on parents
matchingmatching Fixed Influence Method (FI): P=.9Fixed Influence Method (FI): P=.9 Initial Probability Method (AP): P= y+cInitial Probability Method (AP): P= y+c Parent Probability Method (PP): P= x+cParent Probability Method (PP): P= x+c
![Page 14: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/14.jpg)
Probability DistributionProbability Distribution
Probability Distribution for mapping between C and C’
![Page 15: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/15.jpg)
Combining InfluencesCombining Influences
We assume that the parents are We assume that the parents are
conditionally independentconditionally independent
P[C|A,B] = P[C|A] x P[C|B]P[C|A,B] = P[C|A] x P[C|B]
Fix of this for future workFix of this for future work
![Page 16: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/16.jpg)
ResultsResults
2 Sets of 11 and 19 nodes2 Sets of 11 and 19 nodes Predicate matching was manualPredicate matching was manual Thresholds were .85 and .15Thresholds were .85 and .15
![Page 17: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/17.jpg)
ResultsResults
![Page 18: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/18.jpg)
StrengthsStrengths
Innovative researchInnovative research
Published at ISWCPublished at ISWC
Mathematically orientedMathematically oriented
![Page 19: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/19.jpg)
WeaknessesWeaknesses
Lots of typosLots of typos
No comparison with current methodsNo comparison with current methods
Little literature researchLittle literature research
Could use better explanation of basic Could use better explanation of basic
conceptsconcepts
![Page 20: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/20.jpg)
Future WorkFuture Work
Handling conditionally dependency of Handling conditionally dependency of
parent nodesparent nodes
Handling of matching predicatesHandling of matching predicates
Automatic pruning and building of Automatic pruning and building of
the networkthe network
![Page 21: OMEN: A Probabilistic Ontology Mapping Tool](https://reader035.fdocuments.net/reader035/viewer/2022062315/56815249550346895dc085ac/html5/thumbnails/21.jpg)
??