Ontology Evolution

1
Ontology Evolution

• York Sure
• SDK Cluster Meeting
• Karlsruhe, 6.4.2004
• With contributions from Ljiljana Stojanovic and
  Peter Haase

2
Agenda

• Work done so far KAON evolution features by
  Ljiljana Stojanovic (FZI)
• Future work in SEKT

3
Motivation

• Ontology evolution is the timely adaptation of an
  ontology and the consistent propagation of the
  changes to dependent artifacts
• The variety of causes and consequences of the
  ontology changes makes ontology evolution a very
  complicated operation
• What are the problems?
• Complexity - ontology data models are rich
• Dependencies - ontologies often reuse and extend
  other ontologies
• Physical distribution - ontology development is
  a de-centralized and collaborative process
• What is the solution?
• The resolution of these problems in a systematic
  manner

4
Ontology Changes
Concept
Property
Symmetric
Transitive
Inverse
O
Concept hierarchy
Property hierarchy
domain
range
Min cardinality
-
Max cardinality

Instance
instcon
IP
instprop
literals
5
Ontology Consistency

• An ontology is consistent if it satisfies all
  invariants of the ontology model
• Invariants are constraints that must hold in
  every stable state of an ontology

6
Ontology Evolution - requirements

• Functional requirement
• enable the handling of ontology changes
• ensure the consistency of the underlying ontology
  and all dependent artefacts, e.g. instances
• Guidance requirement
• support the user to manage changes more easily
• Refinement requirement
• offer advice to the user for continual ontology
  refinement

7
Ontology Evolution - Process
Core component
8
Semantics of change

• Enables resolution of changes in a systematic
  manner by ensuring consistency of the whole
  ontology

X
9
Propagation Evolution strategies

• Resolution points
• how to handle orphaned concepts
• how to handle orphaned properties
• how to propagate properties to the concept whose
  parent changes
• what constitutes a valid domain of a property
• what constitutes a valid range of a property
• ..

- delete property domain - propagate to the
superconcepts - propagate to the subconcepts
10
Capturing Change Discovery

• Explicit request by the user
• Implicit request through learning

- Structure-driven exploits a set of heuristics
to improve an ontology based on the analysis of
the ontology structure
- Data-driven - detects the changes based on the
analysis of the ontology instances
- Usage-driven takes into account the usage of
the ontology
If no instance of a concept C use any of the
properties defined for C, but only properties
inherited from the parent concept, we can make an
assumption that C is not necessary.
If all subconcepts have the same property, the
property may be moved to the parent concept
By tracking when entity has last been retrieved
by a query, it may be possible to discover that
some entities are out of date
11
Some Directions
Nodes
Ontologies
12
Agenda

• Future work in SEKT
• Evolution strategies for OWL ontologies
• Ontology versioning
• Mapping adaptation for evolving ontologies
• Change Discovery
• General SEKT Picture

13
Evolution Strategies for OWL Ontologies

• Extension of existing evolution facilities
• Challenge
• More complex ontology model
• Different notion of consistency(e.g. semantics
  of constraints)
• Various fragments of OWL (DL, Lite, DLP, )
• Goal Support for evolution strategies in
  Wonderweb OWL API

14
Ontology Versioning

• Ensure consistency of versions
• Manage versions of dependent and distributed
  ontologies
• Manage relations between versions in terms of
• Change operations (Evolution Log)
• Mappings
• Multi-Version Reasoning (with VUA)

15
Mapping Adaptation for Evolving Ontologies

• Mappings to transform data between different
  representations
• Changes in ontologies need to be reflected in
  mappings
• Inconsistencies in mappings need to be detected
  and updated
• Potential Scenarios
• Local mappings between peers in P2P system
• Data integration of heterogeneous data sources

16
Change Discovery

• Explicit request by the user
• Implicit request through learning

- Structure-driven exploits a set of heuristics
to improve an ontology based on the analysis of
the ontology structure
- Data-driven - detects the changes based on the
analysis of the ontology instances
Use Human Language Technology and Data/Text
Mining to support these tasks!
- Usage-driven takes into account the usage of
the ontology
17
General SEKT Picture

• Core Technologies
• Human Language Technology
• Data/Text Mining
• Ontology Metadata Management

18
KAON Evolution Features

• DEMO!

19
Semantics of change

• Enables resolution of changes in a systematic
  manner by ensuring consistency of the whole
  ontology

X
20
Semantics of change
21
Semantics of change
Requested Changes
Requested Derived Changes
22
Evolution strategies

• Resolution points
• how to handle orphaned concepts
• how to handle orphaned properties
• how to propagate properties to the concept whose
  parent changes
• what constitutes a valid domain of a property
• what constitutes a valid range of a property
• ..

- delete property domain - propagate to the
superconcepts - propagate to the subconcepts
23
(No Transcript)
24
The End.

• Thank You!

25
Usage-driven Change Discovery
Request
Response
E
Domain Ontology
M
Log
Log Ontology
A
P
26
Usage-driven Change Discovery

• Tasks
• Discovery of problems
• find the places in the ontology that do not
  fulfil the needs of end-users
• define the corresponding measures
• Generation of changes
• map the problem into the set of ontology changes
• define the interpretation of the extremes

• Users activities
• Querying
• Browsing