[
Lists Home |
Date Index |
Thread Index
]
- From: Joseph Kesselman/Watson/IBM <keshlam@us.ibm.com>
- To: xml-dev@lists.xml.org
- Date: Wed, 18 Oct 2000 12:48:10 -0400
>start from the
>dog and do a search of everything connected to it to try to find the
>name. AI comes into the heuristics in navigating around a database of
>information.
Every successfully solved "AI" problem eventually turns out to be a
pattern-matching problem. The challenge has been finding faster and more
robust pattern engines.
The "semantic web" is definitely a pattern-matching problem. Whether you
call it AI or not is just a matter of how you want to think about it, how
ambitious you expect to get with it... and what subtexts you do or don't
attach to the term AI.
In the most general form, it's definitely an AI problem writ large. "Buffy
DuBray Beagle of Britt's Farm" may be called "Cuddles" by her owner, "Buff"
by the neighbors, "Pooch" by the mailman (who uses that name as a generic
term for any canine significantly larger than a chiuaua),....How do you
conclude that you have found the name that the user wants to use, when some
people/applications will recognize one but not another? When do you decide
the search is becoming overly expensive? How do you prevent repeatedly
searching portions of the mesh?
...But above and beyond, there's still one name left over
And that is the name that you never will guess
A name that no human research can discover
But the cat himself knows, and will never confess
-- Old Possum's Book of Practical Cats; pls. pardon any glitches
in the quote.
Maybe the Semantic Web isn't supposed to be that ambitious, and will be
more practical thereby. But until someone presents a real architectural
diagram for it, we really can't tell what it is or whether it's
practical... and frankly, until that information is presented, I don't
think there's any great value in trying to anticipate and draw inferences
re its impact on any of the other work being done.
If there's a real study group running on this, trying to drive it to use
cases and derive implications and dependencies, great. When they produce a
requirements document, use case scenarios, and the like I'll be glad to
consider them.
Until then, I think the answer is to simply continue to make XML as robust
and general as we can, and trust that they will find a way to build
whatever-it-is on top of that basic rich-syntax framework.
______________________________________
Joe Kesselman / IBM Research
|
