Re: SUO: Re: Focus and Volume
Pat,
It's possible that Doug's views have changed. Or that he himself has
alternate views. In any case, I'd suggest that any assertion that "only
bottom up is good" or "only top down is good" are false. The most sensible
approach (which does seem self-evident to me) is that one does both. Make
high level distinctions as best one can a priori and then test them against
modelling tasks in common sense reality or particular applications. That's
what my group does every day.
Let me pose a pointed question in turn. If you really believe that
there shouldn't be an upper ontology, or that it should only contain Entity
and Relation, are you not in fundamental disagreement with the Scope and
Purpose defined for this group?
Let me however leave that as a rhetorical question. I think in fact we
can avoid further haranguing over lofty and general goals if we focus on
actual substance. Would you support an effort to create simply a catalogue
of theories, along the lines of your paper "A catalogue of temporal
theories"? If so, I think our efforts on the SUO "on the ground", as it
were, will be identical. We'll just be trying to formalize models of time,
spatial relations, human organizations etc. It will just happen to be the
case that some folks will be trying to make those theories consistent with
each other.
Adam
At 09:09 PM 5/14/2001 -0500, pat hayes wrote:
>>Pat,
>> It would be incorrect to place me at the opposite end of the spectrum
>> from Doug. I'd summarize his position as that an upper ontology is
>> essential and relevant
>
>The last time I met Doug (about 6 months ago) I talked to him about the
>SUO, briefly. I asked him whether it wouldnt make sense to simply remove
>the 'upper ontology' from CYC altogether, and have an inheritance
>hierarchy which simply had a large number of top nodes (which would be,
>roughly, the tips of the upper ontology classification.) He agreed, and
>said that he had in fact seriously suggested making that change some years
>earlier, but that his customer base all protested that they preferred to
>have a single top node. He shrugged, with the implication that if thats
>what they want, he was willing to let them keep it. So I think that your
>and Doug's viewpoints are not quite identical; he thinks the upper levels
>of the ontology are irrelevant and distracting.
>
>But let me be the straw man. I will stake out a position and defend it:
>the upper ontology is not irrelevant, but almost any upper ontology does
>more harm than good. That is, almost any ontology could be constructed
>more efficiently and with less work, if the highest levels were stripped
>out or ignored, and then perhaps inserted after the fact to 'fit' the
>middle-level concepts. The reason being that it is essentially impossible
>to think of the necessary distinctions ab initio, but that one must
>discover them by building nontrivial axiomatic theories at the 'middle'
>levels first, and thereby discovering the conceptual distinctions that
>provide the best representational leverage. Most of the high-level debates
>on this list that have been of real ontological substance (eg the
>continuant/occurrent debate, the relevance of 4-d spatiotemporal
>frameworks) have been on issues that have already been distilled from
>earlier middle-level work (in that case, an entire literature), and most
>of those that have not been so distilled (eg almost everything that is in
>any way associated with the intials CSP) have been an entire waste of time
>and bandwidth. Meanwhile, the merged ontology consists of an accumulation
>of middle-level theories being hung onto a framework which was provided a
>priori rather than being abstracted from the theories themselves (in
>direct opposition to the methodology of the author of that heirarchy, by
>the way) and shows the resultant conceptual strain at almost every point.
>
>I might add that this methodology is supported to some extent by empirical
>work on concept mining, where a key strategy in identifying useful
>conceptual dimensions is to find the ways that concepts can be teased
>apart, ie how they differ as well as the ways they are similar. The key
>question to ask is, given three concepts A, B and C: in what way would B
>be closest to A and furthest from C? The point being that this kind of
>analysis requires a level of detail which only becomes apparent when one
>has a reasonable amount of particular detailed knowledge to apply it to.
>
>>but that there are many "good enough" upper ontologies that one might
>>construct and all would be equally useful.
>
>Well, in the sense that zero utility is equally useful as zero, I guess
>that you and Doug might agree.
>
>>The point is to pick one (or generate one) and move on, as agonizing over
>>the "one true ontology" is pointless. I agree with Doug whole-heartedly.
>
>No, the point is to move on. The issue is not the one true ontology, but
>the utility of ANY upper ontology. If the upper ontology is really not
>that important, then why are we (you) wasting so much time building one?
>What utility is shared by *all* upper ontologies?
>
>Look, to make the point more, er, pointedly. Here is a proposal for an
>upper ontology. It is a tree with two levels. The root is called "entity",
>and its children are all the middle-level concepts that anyone wants to
>use. To add a new such concept is easy: it is simply a sibling of all the
>others. Call this the SHUO, the SHallow Upper Ontology. Lets take that as
>a starting point and at some point, when we have a reasonably large SHUO,
>try applying the techniques outlined on Sowa's website to infer a suitable
>lattice-ised version of it.
>
>Now, that might be some actual use. Even for those who feel that a
>philosophical perspective should provide some useful structure for an UO,
>this would be an interesting experiment. If the automatic techniques
>rediscover some well-known philosphical distinction, they can claim some
>empirical success. If they do not, some may feel that this is a diagnostic
>for more axiomatic work to be done, while others may feel that their
>disdain for mere philosophizing has been vindicated. But the result is
>more likely to be useful, and is MUCH less likely to be contentious, in
>either case. We could make a SHUO-ish version of the merged ontology just
>by stripping out all the upper levels, by the way.
>
>Pat
>
>>
>>At 07:07 PM 5/14/2001 -0500, pat hayes wrote:
>>
>>>>Jon Awbrey wrote:
>>>> > .......
>>>>........
>>>>......
>>>>......
>>>>.......
>>>>........
>>>>.........
>>>> >
>>>> > 3. Square One is that everybody who makes a contribution to the
>>>> > main list probably does so because they honestly believe that
>>>> > it has a bearing on what a State of the Art Third Millennium
>>>> > Ontology System ought to look like, even if that makes some
>>>> > State of the Art 1979 folks uncomfortable with the amount of
>>>> > work and the number of different expertises that would really
>>>> > be required to avoid cranking out yet another standard that
>>>> > all innovative people will have to work against from Day One.
>>>>
>>>>
>>>>There's a spectrum of views on
>>>>the extent to which the SUO should be
>>>>pushing the state of the art as opposed to more simply
>>>>reflecting current accepted/best practice.
>>>
>>>It isnt clear to me that there really IS a currently accepted or best
>>>practice in this area. We would do well to remember that there really
>>>are only a few upper ontologies in existence, and many of their authors
>>>are contributors to this list (and the other major author, Doug Lenat,
>>>is firmly of the opinion that the upper levels of a large-scale ontology
>>>are largely irrelevant.)
>>>
>>>>In my opinion, a shared statement on where we are along this spectrum
>>>>would give us a metric which would identify a good fraction of all
>>>>previously posted paragraphs as not relevant (to the group per se).
>>>
>>>I would suggest that a more useful spectrum along which to place
>>>opinions would be the one that has Doug at one end and probably Adam at
>>>the other, which estimates the likely relevance of upper levels of the
>>>ontology to anything said at lower levels. Doug's view, if I can capture
>>>it well enough in a brief sentence, is that the inheritance structure of
>>>any ontology only becomes important when the density of axioms becomes
>>>reasonably high, since until there are significant numbers of things to
>>>say that are likely to be usefully inherited downwards, the heirarchy is
>>>really no more than decoration. Put into less contentious and more
>>>functional terms, this might be paraphrased as the methodological
>>>requirement that any high-level distinction be justified in terms of
>>>some assertions that are likely to have inheritable downward
>>>consequences, ie that something useful can be inferred about some thing
>>>or category X by virtue of it being placed under the high-level
>>>category. Doug's opinion can then be derived from his extensive
>>>experience as the observation that very little can be usefully said at
>>>the very high levels which is likely to have any useful consequences at
>>>any lower levels, that could not be said better or more easily at those
>>>lower levels directly.
>>>The other end of this opinion spectrum holds that getting the upper
>>>levels 'right' in some sense (there may be several senses involved, by
>>>the way) is a prerequisite, or at any rate a desideratum, for creating
>>>an ontology with well-structured lower levels. This is the more popular
>>>opinion - some seem to take it as almost axiomatic - but it seems to me
>>>that this case is the harder one to make, if a case needs to be made. I
>>>would suggest that a more useful and productive debating topic than the
>>>ones that have been cluttering this list so far might arise if those who
>>>feel that an upper ontology is obviously useful, could provide arguments
>>>for this position. At the very least, this would be a start at exposing
>>>the various assumptions and goals that drive the SUO effort, and might
>>>help to clarify how these various opinions relate to each other. Since
>>>we often seem to be trying to pull the effort in different directions,
>>>this might be worth taking a little time over.
>>>
>>>Pat Hayes
>>>
>>>---------------------------------------------------------------------
>>>IHMC (850)434 8903 home
>>>40 South Alcaniz St. (850)202 4416 office
>>>Pensacola, FL 32501 (850)202 4440 fax
>>>phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
>>
>>Adam Pease
>>Teknowledge
>>(650) 424-0500 x571
>
>---------------------------------------------------------------------
>IHMC (850)434 8903 home
>40 South Alcaniz St. (850)202 4416 office
>Pensacola, FL 32501 (850)202 4440 fax
>phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
Adam Pease
Teknowledge
(650) 424-0500 x571