Hans and Kingsley, (01)
That point is critical: (02)
> Remember that applications are not the only context elements/dimensions to
> which data is coupled. Nor is it easy to decouple data from such context
> elements. I believe that's the point John Sowa was making earlier in this
> thread. And applications/services often include a lot of context information
> (like business rules and domain knowledge) that is not explicitly
> represented in the data, making correct interpretation of the "naked" data
> problematic. (03)
I didn't discuss these issues in this thread, but they are fundamental
to anything represented in ordinary language -- and that includes
everything that anybody experiences, thinks about, or works on. (04)
For NLP, taking a word, phrase, or sentence "out of context" is usually
misleading and often hopelessly wrong. People working on NLP recognized
those issues back in the 1950s. After 60 years of R & D, some special
cases can be handled, but there are no workable general solutions. (05)
Providing namespaces with unique identifiers and precise definitions
can be useful for many special cases. NLP researchers have been doing
that on single documents for decades -- but with limited success. To
claim that a partial solution on a small scale will magically work
when applied to the WWW is wishful thinking. (06)
KI
> Data issues are *no* longer nascent and somewhat lost is marketing babble
> spewed by DBMS vendors. My data and applications that help me work with my
> data are two distinct things. (07)
I agree with the second sentence. But data issues had been analyzed
by linguists and logicians since Aristotle. They certainly weren't
"nascent" in the 1960s. And I consider the SW hype as misleading
and misguided as any marketing babble. (08)
KI
> The need to loosely couple Data and Applications (including DBMS
> engines) *wasn't* so clear in the past. (09)
Don't forget the punched-card machines. By the 1950s, there had been
half a century of experience with "loosely coupled" decks of cards
that were processed by a wide variety of different applications. (010)
Every large business of any kind had huge volumes of cards, which
they rapidly computerized in the '50s and '60s. The "loose coupling"
of cards inspired them to design DBs with a similar level of loose
coupling. That was in 1963 for GE's network DB, and 1966 for IBMs
hierarchical IMS. Each record in IMS was a "virtual card". (011)
In fact, they used to mail those cards to customers, who wrote more
information on them. Toll collectors would hand punched cards to
the drivers -- with the warning "Do not bend, fold, or mutilate". (012)
The decks of card were organized in columns, which were printed
as tables or spreadsheets. In 1969, Ted Codd recognized that
those tables could be represented as relations in logic. That
led to a very clear decoupling. (013)
Meanwhile, the AI community had been working on logic-based
representations since the 1950s. For example, see John McCarthy's
"Basis for a mathematical theory of computation" from 1961: (014)
http://www-formal.stanford.edu/jmc/basis1.pdf (015)
That paper talks about n-tuples for representing functions and
relations. McCarthy also used his "conditional expressions",
which he introduced into LISP in the late 1950s. He also worked
with the Algol committee and proposed if-then-else as the
"syntactic sugar" for conditionals. (016)
By the end of the '70s, the combination of the practical and
theoretical experience led to a huge literature about the many
practical and theoretical issues with DBs and KBs. (017)
In comparison, I would call the SW hype naive, provincial, and
based on wishful thinking that was untested against reality. (018)
John (019)
_________________________________________________________________
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 (020)
|