Dear John, (01)
> To evaluate an ontology, we need to distinguish the requirements for
> the upper-level ontologies, the middle levels, and the low-level
> microtheories. Philosophy dominates the upper levels. Very narrow,
> very specific problems dominate the lower levels. And terminologies
> for various application domains dominate the middle levels.
> The criteria for evaluating an ontology are different for each level.
> Following is an excerpt from a note I sent to Ontolog Forum in response
> to Chris Partridge.
> After the excerpt, I added more discussion about the criteria at
> different levels. (02)
MW: On which I will comment.
> > When a top ontology has been introduced, the turnaround has reduced
> > significantly. The top ontology gives a framework for focusing the
> > discussion.
> I strongly agree. But that top level must be very underspecified, and
> the specific ontologies for different purposes from different points of
> view will inevitably be mutually inconsistent. (03)
MW: I don't know what you mean by "underspecified". If you mean it should
not be over constrained, I would agree, but that is true of any ontology at
any level. In my experience it is the most common cause of problems. So when
you put a constraint into your ontology you had better be sure that it is
not just true in this context, but always and everywhere.
> Furthermore, there is no such thing as one ideal top level.
> A 4D ontology is great for many purposes, and a 3D ontology is better
> for others.
> For interoperability, you do not need agreement at the top level, which
> is only "a framework for focusing the discussion." (04)
MW: It is a lot more than that. Especially when the mid-level ontology is
being developed by a large team (tens to hundreds).
MW: When you have a team approach you need to have a development paradigm
that consists of commitments, rules, and choices that as far as possible,
and this is what the upper ontology needs to provide. It defines the
paradigm so that the mid and lower level ontologies are developed
consistently. The key requirements are that:
1. It allows you to say anything that is valid.
2. It provides just one way to do it. (this means it might be over specified
in some areas to eliminate options).
3. It is sufficient, clear and accessible not to require guru intervention.
4. It is accessible at different levels of expertise. (05)
You then need to have a quality approach that means that if the paradigm
does not work, you improve the paradigm. This is important not least because
as far as I know, no one has managed to meet all of those requirements yet. (06)
> And you do not need
> agreement at the lowest levels, which are extremely problem specific. (07)
MW: Yes. Where there are conflicts between theories, you just need to make
sure that you know which one you are using when.
> Where you do need agreement is at the middle level of the words that
> people use to talk about a subject. (08)
MW: And as Chris P pointed out, that is what having an upper ontology
> That is why I believe that
> Schema.org (and the GoodRelations ontology) have hit the "sweet spot"
> of specifying a useful, underspecified middle level. (010)
> Charles Sanders Peirce stated some fundamental principles, which I
> believe are essential for guiding ontology design and use:
> > It is easy to speak with precision upon a general theme.
> > Only, one must commonly surrender all ambition to be certain.
> > It is equally easy to be certain. One has only to be sufficiently
> > It is not so difficult to be pretty precise and fairly certain at
> > about a very narrow subject.
> In other words, you can do your precise reasoning at the lowest levels
> (about very narrow subjects). Your upper level (general themes) can be
> precise, but only as one of many possible frameworks or guidelines.
> For the middle levels, it's easy to be certain if you keep them
> "sufficiently vague." (011)
MW: I don't think I agree with you entirely. I agree that there are multiple
possible upper ontologies, but it is not a virtue to be vague in the mid
level. The mid-level defines the common language. If the shared
understanding is not sufficiently precise, inconsistency will result in its
use, and that is expensive, so vagueness is something I would want to
minimise (I don't pretend you can eliminate it).
> Many people have been unhappy about Schema.org because of their lack of
> axioms. But that's necessary to keep it "sufficiently vague" to
> support interoperability among an open-ended variety of systems.
> GoodRelations, which has been adopted by Schema.org, is another example
> of a good, "sufficiently vague" middle level. (012)
MW: It is not the axioms as such that are important, rather that they pick
out the same things where ever they are used, so it is more about clarity of
the intended interpretation.
> Fundamental criterion for evaluating a middle level: Consistency with
> a variety of different upper levels and the ability to support
> interoperability among lower levels designed for different purposes. (013)
MW: Creating a mid level that is consistent with multiple upper levels where
the upper levels are inconsistent with each other (e.g. 3D and 4D) then you
have a significant challenge, and not likely to happen by accident. (014)
Tel: +44 1489 880185
Mobile: +44 750 3385279
This email originates from Information Junction Ltd. Registered in England
and Wales No. 6632177.
Registered office: 2 Brookside, Meadow Way, Letchworth Garden City,
Hertfordshire, SG6 3JE. (017)
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2013/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2013
Community Portal: http://ontolog.cim3.net/wiki/ (018)