On 9/11/2010 4:18 PM, David Eddy wrote:
> What I thought I was looking for in these ontology discussions was
> the ability to determine which of 35+ potential meanings a cryptic
> "term" means within a piece of software or documentation. (01)
The task of analyzing a document and determining which sense is
appropriate for each word is still a problem that has only been
partially solved. But for those cases where a solution can be
found, it is necessary to have a precise method for recording
that solution. (02)
> So what is the precise logic in ontologies actually useful for? (03)
Logic serves as the notation for representing all the precisely
defined word senses -- it provides one term or expression in logic
for each of those word senses. It also serves as a system for
reasoning about statements that use precisely defined terms. (04)
> To me this is "simply" a systems maintenance issue. If you know
> what the system is doing & what data it manipulates, then it's
> (at least theoretically) easier to interoperate with other systems.
>
> How do ontologies make this—VERY expensive—process easier? (05)
That is a good question. Although I advocate the use of formal
ontologies for subjects that are so well defined that they can
support precise reasoning, I admit that logic is not magic. (06)
For many of the tasks in LOD, the source information is never
precisely stated. Using a precise formal logic as the target
for interpreting very squishy source documents is overkill. (07)
That is one variant of the GIGO principle: If your source
documents have a high garbage quotient, any translation to
logic may look precise, but what it says so precisely will
still be garbage. (08)
> In a chaotic environment (e.g a collection of 100s/1000s of
> applications written & maintained over decades for constantly
> evolving business needs) how does applying organized logic
>single term, single meaning) help with this situation? (09)
An approach that starts by analyzing the NL documents would
fail by the GIGO principle. But it is possible to start with
the implemented programs and databases and generate a precise
translation to logic of what those systems *actually* do
(which may have no similarity to what anybody says they do). (010)
Anything implemented on a digital computer is formal from
the very beginning -- it is not possible to write an "informal"
program. However, we all know that many of those very precise
programs do not correspond to the vague ideas of the people
who use them or the managers who ordered them to be written. (011)
But as a matter of fact, a colleague of mine, Arun Majumdar
solved a problem very much like the one you describe, and he
did it in the way I just summarized: use logic to analyze
the programs, and then check to see how or whether the much
less precise NL documents corresponded to the precise logic. (012)
For a brief summary of that approach, see slides 91 to 98
of the following: (013)
http://www.jfsowa.com/talks/iss.pdf (014)
John (015)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (016)
|