Ed, Pat, and Toby, (01)
This thread gone far beyond my original note, which was not
about apples, oranges, or RDF. It was about human nature. (02)
The point I wanted to make is that telling people to observe
standards is as useful as telling them to be good and to
avoid sinning. That would achieve about a 4.13% compliance,
which is the level of compliance to W3C recommendations. (03)
A logo that certifies that a web page is compliant is more
convincing -- about the same level as an oath to defend the
Constitution in front of a chief justice of the Supreme Court.
That can achieve a 50% level of compliance. (04)
EB> The big issue for the proponents of "ontology repositories" is
> not whether the proffered ontologies are valid OWL/RDF/CLIF. It
> is: How do you decide whether a proffered ontology is "good"?
> And how do you resolve discrepancies when trying to enrich your
> overall knowledge base? Is the new information wrong? or the
> old information? Or does the new ontology just reflect a
> viewpoint for which the reconciliation with other viewpoints
> is not yet fully developed in the community? (05)
I agree. (06)
Syntactic validity, which is easy to check by a parser, is also
easy to enforce by a parser. I agree with Ed's reasons why
the many browsers currently in use don't enforce standards. (07)
But semantic validity cannot be enforced by a parser. In fact,
there is no formal test that can determine whether a given set
of axioms about some real world domain accurately reflect the
facts about that domain. (08)
EB>> The big issue for the proponents of "ontology repositories"
>> is not whether the proffered ontologies are valid OWL/RDF/CLIF.
>> It is: How do you decide whether a proffered ontology is "good"? (09)
PH> You have to trust the source, the publisher. I agree that we
> have no way to compute such trust at the present, but Im sure it
> will evolve. (010)
Perhaps. But even if an ontology is very good and widely trusted,
there is no way to determine whether some person or program that
tags some data with a label XYZ intends it with exactly the same
definition as the person or program that defined the XYZ tag. (011)
PH> ... bioinformatics is very concerned with such issues, and is
> the community most likely to create standards or at least methods
> that might lead to standards for handling such questions. Again,
> all this is active work, being done right now by people for whom
> the results matter. It is no longer armchair theorizing. (012)
Yes, but even in bioinformatics, the agreement among different
highly trained people who have to tag data is low. Differences
of about 30% are typical, and a discrepancy among taggers of 5%
is considered exceptionally good and is rarely achieved. (013)
TC> Surely these issues will be solved by folksonomies of expertise
> and authority, rather than the pay-to-play agencies.
>
> Come to think of it, most of the ontologies will be folksonomy
> based, too. (014)
The folksonomy process is based on the same kind of consensus that
has created and adapted the informal meanings of words in natural
languages. The only difference is the speed that is possible with
the WWW. (015)
For finding movies, restaurants, or job opportunities, that method
is quite good. But you can't depend on a folksonomy to define the
terms for designing an airplane or prescribing the drugs and therapy
for a cancer patient. (016)
The informal terms (of NLs or folksonomies) are the ones that
people actually use in their informal thought and language. But
we need formal definitions to support long chains of reasoning in
computer systems. And we need to map those formal terms to and
from the informal terms that people actually use. (017)
John (018)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (019)
|