|
John,
Thanks for your input. As with Pat's, I only wish it had come sooner. It also clearly shows some areas in which we are miscomminicating to those who did not participate in the summit. That's a serious issue for a Communique.
I reply below to some that you raise newly and some that you share with Pat, that are not touched on in my prior response to Pat.
We will work to correct the flaws in the Communique.
On Apr 21, 2013 10:56 PM, "John F Sowa" <sowa@xxxxxxxxxxx> wrote:
> On 4/21/2013 12:32 PM, Pat Hayes wrote:
> > First, it is way too long and detailed, more of an essay than a
> > communique. It is hard, I would suggest, for anyone to agree with all of
> > it unequivocally. I feel like I want to nit-pick with the text all over
> > the place.
>
> I sympathize with Pat's concerns.
>
> In particular, I very strongly agree with the following point:
>
> > We do not have enough experience with ontology design and deployment
> > to know what the objective standards of "quality" are, still less how
> > to manage teams to achieve this nonexistent standard.
By explicit decision, we are here addressing the engineering sense of "quality" as the meeting of requirements. For this reason, identifying requirements *for a particular intended usage* and finding corresponding evaluation criteria, as well as supporting metrics and methods and tools, is supposed too be much of the focus. This should includes acknowledgement that only some such will be findable. The observations & requirements are supposed to include discussion of needed work in these areas.
We don't know
> > what are the "activities that need to occur during the phases of a life
> > cycle of an ontology", so to go on record with a detailed, confidently
> > stated account which claims to be normative, is both inappropriate and harmful.
Indeed, we discuss certain major activities and how evaluation against requirements can improve their effectiveness, (along with consequent dependence on other activities). We don't want to or try to recommend a single life cycle, but I increasingly think that all text to that effect is unable to counter the many expectations that many readers, (especially those with lots of experience with older, less order-flexible, more unidirectional life cyle models) attach to the life cycle notion and talk of phases or stages. I suspect that this is disorganizing the document rather than organizing it.
>
> I also believe that the technology is developing so rapidly that the
> "best practices" of today are likely to be obsolete in 5 to 10 years.
Yes, as I responded to Pat, this phrase keeps getting removed in one place and popping back up in another. It is misleading, IMHO.
>
> > "Does the ontology follow best practices; in particular does it implement
> > the upper ontology...." Whoa. Is it "best practice" to even HAVE an upper ontology?
> > That is not clear. Most Web ontologies, for example, are not subsumed under any
> > particular upper ontology. If our communique starts being used to justify managers
> > asking ontologists to conform to an upper ontology, we will have done far more harm than good.
> This is a very important point. Even more important is the question
> "What do you mean by an upper ontology?"
>
> For an "integrating ontology" that is designed to enforce a standard,
> a detailed upper ontology may be essential.
>
> But for an ontology that is intended to support interoperability with
> legacy systems, an upper ontology that it too detailed may actually
> *block* interoperability.
Indeed, this is a point well-discussed during the summit. The use of an upper ontology, if any, is generally a architecture and design decision. *If* such use is identified as a requirement, for a partular ontology usage, *then* ontologies intended for that usage should be evaluated against that requirement. If it is not a requiement, it shouldn't come into play. It is an example here of evalating against identified requirements. That is obviosly not communicated to those not part of summit discussions.
> > we read the almost plaintive remark "Generally, appreciation of the full
> > life cycle of an ontology is not well established within the ontology community."
> > Damn right. In other words, none of this is based in actual reality. It is
> > written as though it comprised observations about the right way to work,
> > but in fact, it is not based on observations about how the work is ACTUALLY done.
The "in other words" is not fair, since both this observation and the recommendations regarding what works are, in fact, based on combined experience. But we are not making either the points or the grounding clear to an outside audience, and that matters very much.
> I'll admit that there are some cases where certain techniques were
> actually implemented and found to be useful. But it's not clear how
> far those techniques can be generalized.
This, too -- the need for better understanding of how particular techniques, eval criteria, metrics, etc., are related to particular usages -- is meant to be an explicit point. It seems too have wandered too far out of focus.
> > You might start with CYC, the granddaddy of all large-scale ontologies.
> > You will find that the process bears almost no relationship to the fantasy
> > you describe here.
>
> That is similar to a suggestion I have made repeatedly, but it has
> usually been ignored. I would much, much rather see detailed case
> studies of systems that have actually worked -- including some that
> have been tried and *failed* -- than read some vague generalities
> mixed with a lot of wishful thinking.
>
> John
Best,
Amanda
> _________________________________________________________________
> Committee Msg Archives: http://ontolog.cim3.net/forum/ontology-summit-advisors/
> Subscriber Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit-advisors/
> Community Files: http://ontolog.cim3.net/file/work/OntologySummit2013/
> Community Discussion: : http://ontolog.cim3.net/forum/ontology-summit/
> Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2013
> Community Portal: http://ontolog.cim3.net/wiki/
On Apr 21, 2013 10:56 PM, "John F Sowa" < sowa@xxxxxxxxxxx> wrote:
On 4/21/2013 12:32 PM, Pat Hayes wrote:
> First, it is way too long and detailed, more of an essay than a
> communique. It is hard, I would suggest, for anyone to agree with all of
> it unequivocally. I feel like I want to nit-pick with the text all over
> the place.
I sympathize with Pat's concerns.
In particular, I very strongly agree with the following point:
> We do not have enough experience with ontology design and deployment
> to know what the objective standards of "quality" are, still less how
> to manage teams to achieve this nonexistent standard. We don't know
> what are the "activities that need to occur during the phases of a life
> cycle of an ontology", so to go on record with a detailed, confidently
> stated account which claims to be normative, is both inappropriate and harmful.
I also believe that the technology is developing so rapidly that the
"best practices" of today are likely to be obsolete in 5 to 10 years.
> "Does the ontology follow best practices; in particular does it implement
> the upper ontology...." Whoa. Is it "best practice" to even HAVE an upper ontology?
> That is not clear. Most Web ontologies, for example, are not subsumed under any
> particular upper ontology. If our communique starts being used to justify managers
> asking ontologists to conform to an upper ontology, we will have done far more harm than good.
This is a very important point. Even more important is the question
"What do you mean by an upper ontology?"
For an "integrating ontology" that is designed to enforce a standard,
a detailed upper ontology may be essential.
But for an ontology that is intended to support interoperability with
legacy systems, an upper ontology that it too detailed may actually
*block* interoperability.
> we read the almost plaintive remark "Generally, appreciation of the full
> life cycle of an ontology is not well established within the ontology community."
> Damn right. In other words, none of this is based in actual reality. It is
> written as though it comprised observations about the right way to work,
> but in fact, it is not based on observations about how the work is ACTUALLY done.
I'll admit that there are some cases where certain techniques were
actually implemented and found to be useful. But it's not clear how
far those techniques can be generalized.
> You might start with CYC, the granddaddy of all large-scale ontologies.
> You will find that the process bears almost no relationship to the fantasy
> you describe here.
That is similar to a suggestion I have made repeatedly, but it has
usually been ignored. I would much, much rather see detailed case
studies of systems that have actually worked -- including some that
have been tried and *failed* -- than read some vague generalities
mixed with a lot of wishful thinking.
John
_________________________________________________________________
Committee Msg Archives: http://ontolog.cim3.net/forum/ontology-summit-advisors/
Subscriber Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit-advisors/
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2013/
Community Discussion: : http://ontolog.cim3.net/forum/ontology-summit/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2013
Community Portal: http://ontolog.cim3.net/wiki/
_________________________________________________________________
Committee Msg Archives: http://ontolog.cim3.net/forum/ontology-summit-advisors/
Subscriber Config:
http://ontolog.cim3.net/mailman/listinfo/ontology-summit-advisors/
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2013/
Community Discussion: : http://ontolog.cim3.net/forum/ontology-summit/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2013
Community Portal: http://ontolog.cim3.net/wiki/ (01)
|