uos-convene
[Top] [All Lists]

[uos-convene] Suggestions for structuring the Tu Mar 14 meeting

To: <uos-convene@xxxxxxxxxxxxxxxx>
From: Dagobert Soergel <dsoergel@xxxxxxx>
Date: Thu, 02 Mar 2006 21:59:47 -0500
Message-id: <7.0.0.16.2.20060302173934.03a6de70@xxxxxxx>
This is a composite of two messages I sent earlier and some thoughts that transpired in the uos-convene phone conference on March 2.

Purpose:  To create a map of the upper ontology landscape at several levels:
  • Broad descriptions of each upper ontology, indicating special domains (such as time) included / excluded, level of detail, level of formalization, and philosophical basis / rationale.
  • Broad comparison, arranged in at least two ways: pair-wise comparisons of the ontologies and an outline of issues / constructs with an overview how each is treated in the different ontologies
  • Detailed mapping that compares at the entity and relationship level.  Again, this could be done pair-wise, giving for each entity and relationship of ontology A the closest corresponding entity or relationship from ontology B, explaining agreement and disagreement as appropriate.  And this should be represented in the alternative organization of taking each entity and each relationship, such as partOf, and racing its treatment in all covered ontologies, pointing out agreements and disagreements.
This map could be used to find a common subset, to select an upper ontology for a given purpose, and for creating a "mix and match" upper ontology by taking pieces from several upper ontologies, providing these pieces fit together (which should be decidable based on the map).  (This is the idea of metadata profiles in XML applied to upper ontologies or ontologies in general.)

A general method to arrive at such a map is as follows:

1       Collect suggested upper ontologies - this is done by inviting the "custodians".  Might include ontologies that deal not with all things but with often-used aspects, such as an ontology of time concepts or space or general process description.

2       Compare and determine differences
2.1              In elements (presence / absence and, more difficult, definition)
2.2              in relationships

3       Try to resolve differences, creating a superstructure that incorporates
        the non-contradictory parts of various schemes
3.1              By adding elements
3.2              By adding relationships

4       Articulate the remaining differences so that they are clearly understood.

Some issues that arise in such an effort are listed below (there are surely more).  These issues should be discussed at the Tuesday meeting with the objective of either finding a tentative solution or outlining a way for finding a solution.  Put differently, the meeting should start to hammer out a statement on the difficulties of making these ontologies interoperable and the methods by which they can be made more interoperable.  This would include the determination of
  • areas where there is no or little overlap with one ontology with any of the others,
  • areas where there may be agreement that one of the ontologies is strongest,
  • areas where the custodians may be willing to adapt,
  • areas where people have strongly held different positions
Then issues of process
  • How does one go about resolving differences?
  • Can one agree on a common formal language (or on a common language for more informal description) to describe the differences that cannot be resolved so at least users of the ontologies know what the differences are?
  • What are next steps?  Can two of the custodians start a pilot project of reconciling some parts of their ontologies?
  • How much effort is involved?
There may be more points to be considered.  Perhaps we could get some statements on these points before the meeting.

Dagobert Soergel
College of Information Studies
University of Maryland
4105 Hornbake Library
College Park, MD 20742-4345
Office: 301-405-2037     Home:  703-823-2840        Mobile: 703-585-2840
OFax:   301-314-9145        HFax: 703-823-6427
dsoergel@xxxxxxx     www.dsoergel.com

<Prev in Thread] Current Thread [Next in Thread>