Ontologydriven Resolution of Semantic Heterogeneities in GDB Conceptual Schemas

1
Ontology-driven Resolution of Semantic
Heterogeneities in GDB Conceptual Schemas

• Guillermo Nudelman Hess
• Prof. Dr. Cirano Iochpe

2
Introduction

• Geographic Information Systems (GIS)
  popularization
• Geographic databases (GDB) project
• Complex
• Repeatable

3
Architecture to integrate schemas
CSF Canonic Syntactic File SCSSF Canonic
Syntactic and Semantic File
4
Issues on the semantic integration

• Heterogeneities to be handled Visser, 1997
• Naming (explanation)
• Synonym
• Homonym
• Acronym
• Structure
• Attributes (relation)
• Associations (relation)
• Taxonomy (categorization)
• Constructors (paradigm)

5
Applying Ontology

• Search and Update algorithm
• Semi-automate
• Associated with similarity matching Cohen, 1998
  techniques.

6
The ontoGeo Ontology

• Geographic domain
• Basic cartography
• Hidrography
• Relief
• Vegetation
• Transport
• Locality

• Features
• Spatial representation
• Object
• Field
• Temporality
• Classes
• Attributes
• Relationships
• Something on network

7
The ontoGeo Ontology

• Construction
• Protégé tool
• RDF/RDFs language
• Knowledge model extended ? Synonyms

8
Search and Update Algorithm

• Semi-automate
• Associated with similarity matching Cohen, 1998
  techniques
• Parameters
• Acceptance threshold
• Analysis Threshold
• Confidence ratio (delta value)

9
(No Transcript)
10
Similarity Matching
Levenshtein
Sim(Cc,Co)WN.SimName(Cc,Co)WA.SimEst(Cc,Co) WH.
SimHier(Cc,Co)WR.SimRel(Cc,Co)
WN,WA, WH and WR are the similarity weights for
each component
SimEst(Cc,Co) (?ni1f(Cci,Coi)xWati)/Nat
Wat 1 (Ca/C)
F(Cc, Co) ? levenshtein function applied over
each attribute Wat ? Attributes weight Ca
Number of classes having this attribute C Total
classes Nat ? Number of attributes of the
conceptual schema
11
Similarity Matching
SimHier(Cc,Co) (?(Hier(Cc,Pc).Wt(c,p))/
NHier(Cc,Pc))
Wt(c,p) (E).(d(p)1).(IC(c) IC(p)) E(p)
d(p)
IC(c) -log((?(1/sup(w))).1/N)
Wt(c,p) ? Weight of each IS-A association E
density (classes in the hierarchy) E(p) parent
node density (number of child nodes) d(p) depth
(in the hierarchy) IC Information content
(amount of information) sup(W) number of
parent nodes N Number of nodes in the hierarchy
SimRel(Cc,Co) (?(Rel(Cc,Co))/Rel(Cc))
12
Experiment

• Geographic ontology with about 170 classes
• Hidrology subset
• Parameters
• Accpetance 0,75 (75)
• Threshold 0,4 (40)
• delta 0,1 (10)

13
Experimental results
14
Conclusions

• Ontology
• Schema unification
• Pattern storage.
• Algorithmic methodology to mediate schema
  integration
• Similarity Matching
• Balance of similarities
• Different schemas ? Different weights.

15
Future work

• Categorize the different types of conceptual
  schemas
• Try other similarity matching methodologies for
  each class of schema
• Balance of the WN, WA, WH and WR parameters,
  depending on the input conceptual schema
• Add spatial relationships to the algorithm
• Algorithm optimization.
• Ontology maintenance

16
Thank you

• Guillermo Nudelman Hess
• hess_at_inf.ufrgs.br

?
17
Example
18
Example similarity marching
19
Example similarity matching

• Lago e arroio
• SimName 0,17
• SimAt 0 (arroio does not have attributes)
• SimHier (10,9)/1 0,9 (only 1 hierarchy)
• SimRel 0 (no aggregation associations)
• WN WA WH WR 0,25
• 0,25(0,17) 0,25(0) 0,25(0) 0,25(0,9) 0,27
• WN 0,4 WA 0,3 WH 0,3 WR 0
• 0,4(0,17) 0,3(0) 0(0) 0,3 (0,9) 0,34

20
Example similarity matching

• Lago e Lago
• SimName 1
• SimAt (1(0,5) 0,2(0,75))/3 0,22
• SimHier (10,9)/1 0,9 (only 1 hierarchy)
• SimRel 0 (no aggregation associations)
• WN WA WH WR 0,25
• 0,25(1) 0,25(0,22) 0,25(0) 0,25(0,9) 0,53
• WN 0,4 WA 0,3 WH 0,3 WR 0
• 0,4(1) 0,3(0,22) 0(0) 0,3 (0,9) 0,74