A Review of Ontology Mapping, Merging, and Integration

1
A Review of Ontology Mapping, Merging, and
Integration

• Presenter Yihong Ding

2
Survey Papers

• Ontology Research and Development Part 2 A
  review of Ontology Mapping and Evolving, Ying
  Ding and Schubert Foo
• Some Issues on Ontology Integration, H. Sofia
  Pinto, A. Gomez-Perez, and Joao P. Martins

3
Ontology Mapping

• Two parties understand each other
• Use the same formal representation
• Share the conceptualization (so the same
  ontology)
• Not easy to let everybody to agree on the same
  ontology for a domain
• The problem of ontology mapping
• Different ontologies on the same domain
• Parties with different ontologies do not
  understand each other

4
Ontology Integration

• Building a new ontology and reusing other
  available ontologies (integration)
• Merging different ontologies into a single one
  that unifies all of them (merging)
• Integration of ontologies into applications (use)

5
Integration

• Resulting ontology can be composed of several
  modules
• Be able to identify regions taken from different
  integrated ontologies

6
Merging

• Hard to identify regions taken from merged
  ontologies
• Knowledge from merged ontologies is homogenized
• Knowledge from one source ontology is scattered
  and mingled with the knowledge that comes from
  other sources

7
Use

• Ontologies should be compatible among themselves
• Issues for compatibility
• Ontological commitments
• Language
• Level of details
• Context
• etc.

8
InfoSleuths reference ontology

• Mapping
• Explicit specified relationships of terms between
  ontologies
• Encapsulated within resource agents
• Resource agent
• Encapsulate information about mapping rules
• Present information in ontologies (reference
  ontologies)
• Reference ontologies
• Represented in OKBC
• Stored in OKBC server
• Ontology agents provide specifications
• To users (for request formulation)
• To resource agents (for mapping)

9
Stanfords ontology algebra

• Mapping
• Established articulations that enables the
  knowledge interoperability
• Executed by ontology algebra
• Ontology algebra
• Operators
• Unary filter, extract
• Binary intersection, union, difference
• Inputs ontology graphs
• Semi-automatic graph mapping
• Domain experts define a variety of fuzzy matching
• Use articulation ontology (abstract mathematical
  entities with some properties)

10
AIFBs formal concept analysis

• Mapping and merging
• Ontology concepts with the same extension
• Executed by FCA-Merge
• FCA-Merge
• Create a concept hierarchy - the concept lattice
  -containing the original concepts based on the
  source ontologies
• Process
• Objects annotated by both ontologies directly
  compute lattice
• Else create annotated objects first.
• Else if cannot annotate use documents as
  artificial objects. I.e., concepts which always
  appear in the same documents are supposed to be
  merged

11
ECAI2000s methods

• Williams Tsatsoulis
• Supervised inductive learning
• Create semantic concept descriptions
• Apply concept clustering algorithm to find
  mapping
• Tamma Bench-Capon
• Name-based matching
• Relate classes in bottom-up and top-down ways
• Priority functions to solve inconsistency
• Human experts adjust priority functions
• Uschold
• Use a global reference ontology

12
ISIs OntoMorph

• Syntactic rewriting
• Pattern-directed rewrite rules
• Concise specification of sentence-level
  transformations based on pattern matching
• Semantic rewriting
• Modulate syntactic rewriting via semantic models
  and logical inference

13
KRAFTs ontology clustering

• Based on the similarities between the concepts
  known to different agents
• Method
• Use a domain ontology describe abstract
  information (global reference)
• Each ontology cluster define certain part of its
  parent ontology
• Name, instance, relation, compound matchers

14
Heterogeneous Database Integration

• A database scheme is a lightweight ontology
• Typical researches
• Batini et.al. (1986), five steps of integrating
  schemata of existing or proposed databases into a
  global, unified schema
• Sheth Kashyap (1992), semantic similarities in
  schema integration
• Palopoli et.al. (2000), two techniques to
  integrate and abstract database schemes

15
Other Ontology Mappings

• Lehmann Cohn (1994)
• Need more specialized concept definitions
• Li (1995)
• Identify attribute similarities using neural
  networks
• Borst Akkermans (1997)
• Resulted mappings could be considered as a new
  ontology

16
Other Ontology Mappings

• Hovy (1998)
• Several heuristic rules to support the merging of
  ontologies
• Weinstein Birmingham (1999)
• Graph mapping use description compatibility
  between elements
• McGuinness et.al. (2000)
• Chimaera system
• Term merging from different knowledge sources
• Noy Musen (2000)
• PROMPT algorithm for Protégé system
• Ontology merging and alignment for OKBC
  compatible format

17
Conclusion

• Depend very much on the inputs of human experts
• Focus on 1-1 mappings
• Further needs n1, 1n, mn mappings
• Ontology mapping can be viewed as the projection
  of the general ontologies from different point of
  views