a view-based methodology for collaborative ontology engineering (vimethcoe)
DESCRIPTION
A View-based Methodology for Collaborative Ontology Engineering (VIMethCOE). Ernesto Jiménez Ruiz Rafael Berlanga Llavorí Temporal Knowledge Bases Group http://krono.act.uji.es Universidad Jaume I de Castellón (Spain). INTRODUCTION. Introduction Methodology Requirements Related Work - PowerPoint PPT PresentationTRANSCRIPT
A View-based Methodology for Collaborative Ontology Engineering (VIMethCOE)
Ernesto Jiménez RuizRafael Berlanga LlavoríTemporal Knowledge Bases Grouphttp://krono.act.uji.esUniversidad Jaume I de Castellón (Spain)
VIMethCOE 2
INTRODUCTION
A collaborative methodology for the development of ontologies, based on Views.– Requirements– Related Work– The Methodology
• Phases• Knowledge Spaces• The View Mechanism
– State of our work, and future tasks
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 3
NEW DIMENSIONS IN THE DEVELOPMENT
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
The work “Ontologies: How can They be Built?” (Sofia Pinto, and João Martins) establishes the need for new methodologies that consider new dimensions in the development.
VIMethCOE 4
HOW TO ACHIEVE A GOOD BALANCE BETWEEN DIMENSIONS?
We want to achieve a highly dynamic, distributed and partially controlled scenario for the development.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Proposed Requirements:– Modularization– Local Adaptation– Knowledge Abstraction– Personal Views– Argumentation and Consensus
VIMethCOE 5
MODULARIZATION
Ontologies can involve several thousands of concepts, and require several experts involving different domains (molecular, genomics, organs, diseases, etc.).
The definition of modules would facilitate several aspects:– Maintenance and validation of the ontology– Local reasoning– Collaboration– Reuse of knowledge.
Introduction
Methodology Requirements
Modularization
Local Adaptation
Knowledge Abstraction
Views
Argumentation and Consensus
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 6
LOCAL ADAPTATION
Each participant must be able to deal with knowledge in a local and private working space.
Making changes and local copies independent from the community’s knowledge.
i.e.: Protégé editor
Introduction
Methodology Requirements
Modularization
Local Adaptation
Knowledge Abstraction
Views
Argumentation and Consensus
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 7
KNOWLEDGE ABSTRACTION
The development of ontologies may involve experts from several areas.
These experts may only have a partial knowledge of the domain.
So they will be able to contribute in the development of only a portion of the ontology.
Introduction
Methodology Requirements
Modularization
Local Adaptation
Knowledge Abstraction
Views
Argumentation and Consensus
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 8
VIEWS (PERSONAL MODULES)Our methodology propose the operation through a view mechanism– User extend views in their local space
Views are defined by the developer.
This mechanism provides… – Knowledge Abstraction– Facilities for Visualization in ontology editors– Reuse of Knowledge
Introduction
Methodology Requirements
Modularization
Local Adaptation
Knowledge Abstraction
Views
Argumentation and Consensus
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 9
ARGUMENTATION ANDCONSENSUS
Developers extend their knowledge in a local space
Changes over the local knowledge may be published
These changes should be evaluated by the community.– Following a Formal or a semi-formal argumentation
model like Ibis.
Introduction
Methodology Requirements
Modularization
Local Adaptation
Knowledge Abstraction
Views
Argumentation and Consensus
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 10
RELATED WORK Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Classic Methodologies
Collaborative Approaches
Web-based approaches
Fulfilment of the Requirements
VIMethCOE 11
CLASSIC METHODOLOGIES Introduction
Methodology Requirements
Related Work
Classic Methodologies
Collaborative Approaches
Web-based Systems
Characteristics
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Cyc, Kactus, Uschold-King’s method, METHONTOLOGY, On-To-Knowledge (OTK), UPON, etc.
They propose a centralized approach to the ontology development.
They neglect collaboration issues.
VIMethCOE 12
COLLABORATIVEAPPROACHES
Co4, DILIGENT, HCOME, Divergence Occurrences Methodology, (KA)2, the OntoEdit system, etc.
They do not propose a complete methodology with different phases
But solutions to carry out an agreed definition of the knowledge (mainly the argumentation).
Introduction
Methodology Requirements
Related Work
Classic Methodologies
Collaborative Approaches
Web-based Systems
Characteristics
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 13
WEB-BASED APPLICATIONS
Ontolingua Server, WebOnto, WebODE, Wiki@nt, (KA)2, etc.
They rely completely on the WWW.
So they provide good frameworks for collaboration.
Introduction
Methodology Requirements
Related Work
Classic Methodologies
Collaborative Approaches
Web-based Systems
Characteristics
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 14
CHARACTERISTICS OFRELATED WORK
MODULARIZATIONLOCAL
ADAPTATIONKNOWLEDGE
ABSTRACTIONVIEWS
ARGUMENT. CONSENSUS
CO4
DILIGENT
HCOME
Div. Occurr.
OntoEdit
Ontolingua
(KA)2
WebOnto
WebODE
Wiki@nt
VIMethCOE
Introduction
Methodology Requirements
Related Work
Classic Methodologies
Collaborative Approaches
Web-based Systems
Characteristics
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 15
BREAK POINT - SUMMARY
I have presented the requirements for collaborative methodologies.
I have reviewed some related work
Next, I’m going to present the characteristics of our methodology
VIMethCOE 16
THE VIMethCOE METHODOLOGY
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Complementary to centralized methodologies
We distinguish 5 different phases:– Requirements– Development– Publication and Argumentation– Evaluation and Maintenance– Application
Overlapped Phases
VIMethCOE 17
REQUIREMENTS PHASE
Defining an initial knowledge– Definition of a Top-level ontology, or reusing
(e.g.: SUMO, DOLCE).– Reusing of Ontologies (NCI, FMA, GO, etc.).
Modularization of this knowledge.– Applying a partitioning algorithm
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Requirements
Development
Publication
Evaluation
Application
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 18
DEVELOPMENT PHASE
Knowledge engineers, ontology engineers, domain experts and final users must take part in this phase.
Each participant will define development views in order to extend them.
They will work in a local and private environment
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Requirements
Development
Publication
Evaluation
Application
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 19
PUBLICATION - ARGUMENTATION
Local adaptations of the knowledge can be published, by means of views.
This published knowledge must be discussed by others developers.
Whenever a consensus is reached, the global ontology must be updated.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Requirements
Development
Publication
Evaluation
Application
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 20
EVALUATION - MAINTENANCE
Checking for Consistency
General Argumentation
Redefinition of the initial modules if the growth of the ontology requires it.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Requirements
Development
Publication
Evaluation
Application
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 21
APPLICATION PHASE
In this phase we define views with an application purpose.
They will represent a complementary knowledge.
These views may present divergences with other views and with the global knowledge.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Requirements
Development
Publication
Evaluation
Application
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 22
KNOWLEDGE SPACES Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
In VIMethCOE we propose the coexistence of several overlapped knowledge spaces:
VIMethCOE 23
KNOWLEDGE SPACESPrivate Space– The working space of developers. – The set of views in development and not published
Public Space – Shared knowledge, which can be used by the community.
Agreed Space – Knowledge that is in consensus. – Composed by the ontology modules, and the agreed views
over them.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 24
KNOWLEDGE SPACESDevelopment Views. – Composed by the set of views that aims of extending the
ontology.Application Views. – Composed by the set of views for a specific application.
Old Versions. – To analyse the evolution of knowledge.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 25
THE VIEW MECHANISM
Allows a collaborative evolution of the ontology with dynamism and distribution
But also enables control over the global knowledge. – Control in the definition of views Abstraction– Control in changes over views Argumentation
Next, I’m going to comment the operation of the proposed mechanism
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 26
THE VIEW HIERARCHY
Views can be defined over modules or over other views (agreed or not). Views are grouped inside a hierarchy depending on their definition and the changes made.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
View Hierarchy
Changes over Views
Argumentation
View Language
Prototype
Conclusions
VIMethCOE 27
SITUATION IN VIEW HIERARCCHY
When users publish their extended views, their situations in the view hierarchy are inferred.
If extensions increases the knowledge or causes reversible changes is-a view or derived view.
• i.e.: union of two views
If the view extensions causes some kind of loss of information or inconsistencies conflict view.
• i.e.: deleting a property of a class
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
View Hierarchy
Changes over Views
Argumentation
View Language
Prototype
Conclusions
VIMethCOE 28
ARGUMENTATION PROCESS
Achieve a consensus is a very important aspect for the Methodology.If consensus: – Public View Agreed View– Global Knowledge is updated
No Consensus– Divergent Alternatives (Conflict)– Alternatives may coexist but some control is
necessary the view hierarchy
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
View Hierarchy
Changes over Views
Argumentation
View Language
Prototype
Conclusions
VIMethCOE 29
DEFINITION OF VIEWS
One of the main characteristics of VIMethCOE is the ability to operate through views.
We have designed and implemented a traversal-based view definition language.
Views consist of:– the union of a set of queries – and a set of inference rules
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
View Hierarchy
Changes over Views
Argumentation
View Language
Prototype
Conclusions
VIMethCOE 30
DEFINITION OF VIEWS
Query definitions are paths over the ontology graph with operators over concepts, properties and instances.– “CPT_II Protein”/{componentOf=“MithocondrialMembrane”}
Inference rules may involve the extraction of concepts, properties and instances that are not explicitly indicated in the views: – Objective: Obtain Closed and Complete views.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
View Hierarchy
Changes over Views
Argumentation
View Language
Prototype
Conclusions
VIMethCOE 31
SOME TESTS
Some tests have been realized with a simple prototype.– A plug-in that connects the semi-structured
database G with the ontology editor Protégé
– Views are defined over small ontologies.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
View Hierarchy
Changes over Views
Argumentation
View Language
Prototype
Conclusions More Information: http://www3.uji.es/~ejimenez/GProtegePlugin/
VIMethCOE 32
CONCLUSIONS
We take into account new dimensions (dynamism, distribution and control) in the ontology development and evolution.
We have propose several Requirement in order to achieve a good balance between the dimensions :
– Modularization– Local Adaptation– Knowledge Abstraction– Personal Views– Argumentation and Consensus
We have presented a View-based Methodology that aims to realize the above requirements
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
VIMethCOE 33
A GOOD STATE OF THE ART, AND A GOOD PROPOSAL?
In this work we have carried out a documentation about the state of art in ontology engineering.
And we have proposed a new work for this state of the art.
But, we have only implemented a simple prototype, so we have got more work to do.
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Work in Progress
Future Work
Some Questions
VIMethCOE 34
FUTURE WORK
Formal Definition of Views:– The kind of views proposed is mainly oriented to
frame-based ontologies.– So it is necessary to refine the definition of views to
take advantage of description logic characteristics.
– Formal definition of modules or views around a concept or a set of concepts.
• i.e.: e-modules (Modularizing OWL ontologies with E-Connections, Bernardo Cuenca Grau. et.al.)
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Work in Progress
Future Work
Some Questions
VIMethCOE 35
FUTURE WORK
Application Scenario– The biomedicine domain is an excellent
scenario for applying this methodology.• Large Ontologies like NCI, FMA, GO, GALEN, etc.• The development of ontologies may involve experts
from several areas– i.e.: Molecular, cellular, tissue, organ, individual and
population Modules
– Health-e-Child Project:: http://www.health-e-child.org/
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Work in Progress
Future Work
Some Questions
VIMethCOE 36
APPLICATION SCENARIO Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Work in Progress
Future Work
Some Questions
VIMethCOE 37
SOME QUESTIONS AND CONTACT
Questions?:
Contact:Ernesto Jiménez Ruiz ([email protected])Rafael Berlanga Llavorí ([email protected])Temporal Knowledge Bases Group http://krono.act.uji.esUniversidad Jaume I de Castellón (Spain)
Introduction
Methodology Requirements
Related Work
Phases of the Methodology
Knowledge Spaces
The View Mechanism
Conclusions
Work in Progress
Future Work
Some Questions