Combining Declarative and Procedural Knowledge to Automate and Represent Ontology Mapping

1
Combining Declarative and Procedural Knowledge to
Automate and Represent Ontology Mapping

• Li Xu
• University of Arizona South

2
Ontology Mapping

• A process whereby two ontologies are semantically
  related at conceptual level, and the source
  ontology instances are transformed into the
  target ontology entities according to those
  semantic relations.
• How to find matches?
• How to represent mappings?

3
About approaches to find matches

• A key conclusion is that an effective approach to
  discover matches between ontology elements
  requires a principled combination of several base
  techniques e.g.
• Ontology terms
• Data types
• Data values
• Ontology structures
• Domain knowledge is useful.
• How to organize and represent the knowledge in an
  application domain?

4
Knowledge Representation

• Formal models of domain discourses
• Keywords for vocabulary terms in ontologies
• Metadata description about heterogeneous data
  between ontology instances

5
About mapping representations

• Mappings are for exchanging instances of a source
  ontology to instances of a target ontology.
• Semantic interoperability is to use logic in
  order to guarantee that, after data are
  transmitted from a sender system to a receiver,
  all implications made by one system had to hold
  and be provable by the other, and that there
  should be a logical equivalence between those
  implicationsfrom discussion in IEEE Standard
  Upper Ontology Working Group

6
Declarative and Procedural Knowledge

• Conceptualize domain knowledge into concepts and
  relationships between concepts.
• Express logically equivalent concepts and
  relationships between ontologies.
• Procedural attachment is used to enforce the
  expressive capability.
• Information retrieval methods applied to retrieve
  keywords
• Data extraction methods applied to match data
  with data patterns
• Query computations issued by procedures that are
  beyond limitations of logicbased languages

7
Domain Model Representation

• Input Ontology
• O (S, A, F)
• Domain Ontology
• O (S, A, P)
• Source-to-Target Mapping Ontology
• O (S, A, F, P)

8
Input Ontologies
Color
Year
Feature
Make
Feature
Make Model
Model
Body Type
Style
Miles
Phone
Mileage
9
Domain Ontology
Make Model
Feature
Color
Model
Make
Accessory
Style
Body Type
Concepts and relationships Light weight
ontologies Incremental development
10
Describing Heterogeneous Metadata and data

• Make matches
• constant
• extract CarMakes case insensitive
• lexicon
• CarMakes case insensitive
• filename "carmakes.dict"
• end
• Mileage matches 8
• constant
• extract "1-9\d0,2k" case insensitive
• extract "1-9\d0,2?,\d3" case
  insensitive
• extract "1-9\d3,6" case insensitive
• keyword "\bmiles\b", "\bmi\.", "\bmi\b",
  "\bmileage\b"
• end

11
Finding Matches
Color
Feature
Body Type
Car
Style
Miles
Source Evaluation
12
Source-to-Target Mapping
Color
Feature
Body Type
Car
Style
Miles
Source Evaluation
13
Another Complex Match
Color
Feature
Body Type
Car
Style
Miles
Source Evaluation
14
Source-to-Target Mapping (Cont.)
Year
Year
Feature
Make
Model
Car
Phone
Mileage
Miles
Source
15
Results

• Experiments showed that domain ontologies
  provided a powerful method to explicitly
  represent heterogeneous metadata and data in
  order to support a composite approach for
  ontology mapping
• Semantic interoperability based on a
  source-to-target mapping were proved to guarantee
  that extensional data transmitted from a source
  ontology can be interpreted correctly in a target
  ontology.

16
Conclusions

• Combining procedural and declarative
  representations provides a promising method to
  systematically and incrementally model domain
  knowledge to automate ontology mapping.
• Combining procedural and declarative
  representation for mappings provides a semantic
  bridge to support semantic interoperability and
  to improve query and inference performance.