Re: [ba-ohs-talk] Organic Growth of Knowledge
Good points. My favorite knowledge manager, InfoSelect, gives me "neural"
searches, but also allows me to ontologize to my heart's content via an
outline and cross-tabs. Something like your suggestion. (01)
I feel that for this effort to succeed, nearly all of the work must be
accomplished automatically, because contemporary organizations do not have a
"religion of knowledge." They don't train employees how to answer the phone,
never mind how to ontologize. (02)
----- Original Message -----
From: "Eric Armstrong" <eric.armstrong@sun.com>
To: <ba-ohs-talk@bootstrap.org>
Sent: Thursday, August 15, 2002 2:43 PM
Subject: Re: [ba-ohs-talk] Organic Growth of Knowledge (03)
Malcolm Dean wrote: (04)
> I applaud your efforts and direction, and I wish you tremendous success. I
> still see no great use for it. Google is the best thing since sliced bread
> for finding information, but it doesn't know about my personal ontology,
or
> that of science fiction writers. (05)
There is an "almost" lurking in there that bears some examination, because
it is in the nature of the "almost" that ontologies make sense. (06)
Google is "almost" the best thing since sliced bread, in that
a) It *is* the best thing since sliced bread if you the query you
are asking reflects the moat popularly-asked question.
(The odds are good, because it is in the nature of a popular
question that many people ask it.) (07)
b) If the question you are asking is *different* from the popular
one, Google isn't much help at all. (08)
Let's see if I can construct a concrete example. (I've encountered quite
a few, but haven't recorded them.) (09)
Let's consider a search on something like "stopping a printing process".
Since most Google users are computer geeks, if you want to know
how to cancel a print job, you'll come up with lots of great hits right
away. (010)
But if you're trying to shut down the newspaper's printing press, you're
going to have a huge collection of false positives to get through before
you find something relevant. (011)
That's where ontologies enter in -- to constrain the search to relevant
information. (012)
Then there is the matter of "kind" of information. Somewhere, you
know there is a tutorial that showed a bit of code for determing
whether the text in a file was XML or HTML. What you want is (013)
"code" that "determines" whether a file is "XML or HTML" that's
in a "tutorial". (014)
But "code", "determines/decides", änd "tutorial" are ontological
concepts that describe the material. The only thing could search
on would be "XML HTML" -- and that's going to deliver one
whale of a bunch of false positives. (015)
Of course, the task of ontological tagging is immense. But hopefully
that analysis is just a little bit persuasive as to how it could hold
sizable benefits. As we begin to define reasonable mechanisms to
allow it, the next step is to figure out how it will get done. (016)