[ontac-forum] RE: A suggestion for ontological discussions at ONTAC meet

["Gary Berg-Cross" <gary.berg-cross@xxxxxxxx>]

Nicolas,

 

John Sowa may have an adequate response to your points and questions, but I have a few ideas that I’ll outline and wander over  to further a discussion regarding your issues with an /“open-ended collection of modules that can be selected, assembled, and tailored for particular tasks or collections of tasks/.”

 

You note some challenges to the approach, for example, “It would be difficult to do (a) within a single global world view because we lack conceptual and relational redundancy to make useful comparisons, claims of similarity, equivalence, etc... “

 

Some work in the DOLCE/ DOLCE+ area seems to mix a loose modular including foundational and core ontologies. I’m drawing from the article “Inflammation Ontology Design Pattern: an  exercise in building a core biomedical ontology with Descriptions and Situations by Aldo GANGEMI, Carola CATENACCI and  Massimo BATTAGLIA  from the DOLCE area (http://www.loa-cnr.it/Papers/Inflammation.pdf ). There are quite a few topics in this article, but one main thrust is to expose the concept of  “Ontology design patterns (ODP).”

 

 The authors explain ODP this way:

 

“the most general notions that are defined in visions (foundational ontologies) and in specific domains (core ontologies) are often richly interrelated, usually in a way that allows domain experts to reuse the same set of relations for most of their modelling work. These highly interconnected fragments of foundational or core ontologies can be called ontology design patterns (ODP), analogously to design patterns used in software engineering ….. These are some desiderata for ODPs:

 

• An ODP is a template for solving a domain modelling problem

• An ODP "extracts" a fragment of a foundational or core ontology, which is its "background”

• An ODP is axiomatized according to the fragment it extracts

• An ODP can be represented in any ontology representation language, although its intuitive and

compact visualization seems an essential requirement

• An ODP can be an element in a partial order, where the ordering relation requires that at least

one of the classes or relations in the ODP are specialized

• An ODP can be intuitively exemplified and catches relevant core notions of domains

• An ODP can be often built from informal or simplified schemes used by domain experts,

together with the support of a foundational ontology and a methodology for domain ontology

analysis, or by specializing existing ODPs

• An ODP can/should be used to describe a best practice of modelling

• An ODP is similar to a DB schema, but an ODP is defined with reference to a foundational

ontology and should have a general character, independently from local design details.

 

This is perhaps a first step toward reusable modules that are customizable for tasks.

 

 

I’m using examples  from the article to illustrate their ideas on modular interaction, but the application of their ideas, in the context of your issues, is my own and not the authors, who may not agree with my inferences and application of their work.  I’m using just the introductory part of the article to outline some aspects of the topic.  The full article and it figures provides the details.

 

Prior discussions have used biological examples and cited the “relations in biomedical ontologies” paper (Barry et al 2005) and the DOLCE work cited above is in this area.  The GANGEMI et al article uses the biomedical “domain” as a topic and this is nice for a test of core ontologies and modules since biomedicine has several activity domains (e.g. clinical, administrative financial), different scientific granularities (e.g. the ones that people like Barry cites molecular,  cell  detail, organs), and  different user requirements for the same service (e.g. provider-oriented,  patient-oriented or administrator views) etc. All of which allows us to see how to, as you say we might  “adjudicate which  redundancies are OK vs. inconsistent/incompatible.”  Their biomedical domain includes concepts (such as HL7 tried to cover) in an electron patient record.  Hence the domain of discourse includes theories (diagnostic contexts), guidelines (diagnostic process models), and information objects (texts).

To cover these we need to expand the foundation and core in substantial ways…..

 

They argue that the ontology needs to talk about “knowledge objects” and develop a Descriptions and Situations ontology (D&S) to support a first-order manipulation of knowledge (or descriptive) objects (example of which are plans, diagnoses, norms, institutions, etc. ) and situations (such as cases, facts, settings, etc.  I would guess that reality based ontologists like Barry might not agree with their constructionist approach to “descriptions and situations”, so this may be part of an interesting future discussion.  Nevertheless their examples provide a concrete feeling of the problem of bridging from biological to clinical ontology and adjudicating the relations, which is the example I think worth considering.

 

I’m taking adjudication  as involving the resolution of how intended meaning can be bound to context recognizing that as GANGEMI et al say,  “representation of intended meaning needs a very flexible and rich set of primitives that can be put within a modular architecture, either across ontologies or across elements of an ontology.”

 

Inflammation is the idea they employ to illustrate this since it is linguistically “ambiguous”, spans multiple contexts/layers of reality, as well as multiple details of granularity.   Granularity condition (e.g. biochemical vs. organic)” is perhaps the easiest way to think of different contexts that provide a meaning for inflammation but this interacts with the tasks that are part of a functional view.    A biologist’s view is different than a provider of a healthcare manager’s, whose tasks are different.  BTW, GANGEMI et al conceptualize the interactions between the clinical and biological domains in an extended discussion, well beyond what I can discuss here, but part of their approach is to recognize different existing core biological and biomedical ontologies supported by a foundational ontology in order to interoperate.  But the fundamental interaction between them  is supported by a “typical soft modularization of ontologies into strata and modules which induces certain choices at the top stratum.  “

 

I think that this starts to get at what you said  we need “ways to specify such modules in terms of:

a) some kind of calculus for selecting, assembling, tailoring, etc...

b) an explicit notion of intended purpose that an extenal agent can

attribute to a module    “

 

Their Figure 6 and its discussion shows an epistemological layering of inflammation-related knowledge. At “bottom” is an inflammation biochemical pathway module description that can be “satisfied by a molecular data structure (thus reified as a situation under the pathway description)”. The molecular data structure at the bottom then becomes a constituent of an “organic data structure, which is reified as a situation under an organic inflammation description”. At top in this simplified example is an “ organic data structure that becomes “a constituent of a clinical data structure, which is reified as a clinical condition under a diagnostic inflammation description. “

 

Finally constituent (“setting”) relations holding between the clinical condition and the lower situations (organic and molecular) are inferable via the background axioms of

D&S. Other axioms provide embedding of lower descriptions into the higher ones.

 

Well I hope this is not too confusing and serves as a context for some of the discussion.

 

Regards,

 

Gary Berg-Cross

EM&I

Work Phone: 703-607-3329
Cell: 908-565-4053
Email: gary.berg-cross@xxxxxxxx

 

 

To:	ONTAC-WG General Discussion <ontac-forum@xxxxxxxxxxxxxx>
From:	Nicolas F Rouquette <nicolas.rouquette@xxxxxxxxxxxx>
Date:	Mon, 21 Nov 2005 11:16:09 -0800
Message-id:	<43821CF9.3030707@xxxxxxxxxxxx>

Gary Berg-Cross wrote:    (01)

> The round of topics exposed by the forum discussions since our first 
> meeting have been interesting, but convergence is hard to judge. To 
> help converge towards a consensus based on a focused exploration of 
> the issues I wonder if we can set up a panel (or two) on the issues at 
> the next ONTAC meeting. If I recall right we will have Doug Lenat for 
> a CYC briefing. That might be an opportunity to ask him about hub 
> versus John Sowa’s idea to:
>
> “/Shift attention from the unsolvable problem
> of building, merging, and coordinating global world views to
> /
>
Ok.    (02)

> /the task of developing an open-ended collection of modules
> that can be selected, assembled, and tailored for particular
> tasks or collections of tasks/.”
>
For that to happen, we need ways to specify such modules in terms of:
a) some kind of calculus for selecting, assembling, tailoring, etc...
b) an explicit notion of intended purpose that an extenal agent can 
attribute to a module    (03)

It would be difficult to do (a) within a single global world view
because we lack conceptual and relational redundancy to make useful 
comparisons,
claims of similarity, equivalence, etc...    (04)

In contrast, having multiple perspectives and points of view yields
redudancies for specifying concepts & their relationships. In this case,
it is natural to develop some kind of calculus to adjucate which 
redundancies
are OK vs. inconsistent/incompatible or undesirable in some way.    (05)

Perhaps in an abuse of the term, the problem faced in ONTAC seems to me 
an issue of meta-ontology:    (06)

- how do we specify the context/scope of a module?    (07)

In a conventional approach, a module might be specified as a set of 
tasks and define the ontological concepts/relations
pertaining to these tasks. Another strategy might focus on the 
ontological concepts/relations in the module which would
provide a vocabulary to define tasks.    .......

_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config: 
http://colab.cim3.net/mailman/listinfo/ontac-forum/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: 
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG    (01)