John,
See below in line.
Sjir Nijssen
Chief Technical Officer
PNA Group
Tel: +31 (0)88-777 0 444
Mob: +31 (0)6-21 510 844
Fax: +31 (0)88-777 0 499
E-mail: sjir.nijssen@xxxxxxxxxxxxx
-------------------------------------------------------
http://www.pna-group.com
-----Oorspronkelijk bericht-----
Van: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] Namens John F Sowa
Verzonden: zondag 20 januari 2013 16:23
Aan: ontolog-forum@xxxxxxxxxxxxxxxx
Onderwerp: Re: [ontolog-forum] Knowledge graphs by Google and Facebook
Leo, William, Matthew, Len, and Kingsley,
This discussion gets into issues of default reasoning and approximate reasoning under uncertainty. The umbrella term is 'nonmonotonic reasoning', which has been extensively analyzed for the past 35 years.
(Ray Reiter published his articles about CWA and OWA in 1978.)
The consensus is that the semantics of *every* version of nonmonotonic reasoning is based on a theory of classical reasoning plus a systematic method for handling exceptions.
Guess what? That's exactly what practical DB administrators have been doing since the US census of 1890 with its punched-card DB.
Leo
> but we need to clarify things in this forum, not confuse people, or
> give them false information.
Absolutely! +1
Leo
> Relational databases are nearly always under the closed world assumption.
> There may be exceptions, but they are few.
That is false information. More precisely, the *semantics* of SQL is defined to be certain under the CWA, but the owners of the data and the people who design and maintain the systems know that errors and missing info are inevitable.
WF
> I have never been able to understand this closed world stuff, except
> as a formal theory that was unnecessary and undesirable. I always
> thought that 'not' in a relational query meant 'not found', rather
> than 'not true' until somebody told me about 'closed worlds.'
'Not found' is the safe assumption to make for *every* DB, relational or otherwise. Telling people that a DB is infallible (i.e., CWA) is misleading and confusing. As Leo said, we should avoid such claims.
MW
> DBMS query engines are generally closed world, but the applications
> that use them are not necessarily closed world but may reinterpret the
> closed world results.
Yes. This is consistent with William's comment. The DBMS engines are guaranteed to be absolutely certain *if and only if* the CWA is true.
But everyone who maintains them provides exception handlers for the inevitable cases in which CWA fails.
Leo
> When the external programming language code surrounding a relational
> database essentially creates its own semantics in procedural code
> (which is extensive, I admit) for a given database, it is not the same
> as saying that the relational model of RDBs is open world. It’s not.
Technically speaking, that is true. But practically speaking, William is right: the distinction between CWA and OWA is a theoretical matter that is used in formal discussions. But the cases when CWA is true are almost nonexistent.
Recommendation: Telling people that SQL uses the CWA is misleading, confusing, and nearly always false. Don't use the terms CWA and OWA outside of research publications.
LY
> To further clarify I would like to add that CWA is necessary to allow
> efficient reasoning which is sound complete and decidable...
Kant, Wittgenstein, and every linguist or lexicographer said that precise definitions are only possible for arbitrary stipulations, such as a mathematical theory. Every naturally occurring concept (e.g., the word senses of every word
in the OED) admits exceptions.
And CWA is irrelevant for efficiency. SQL and Prolog run very efficiently with the OWA.
LY
> This is often a source of confusion about query with SQL vs SPARQL,
> for example.
There is no confusion whatsoever between SQL and SPARQL. Everybody who maintains a DBMS of any kind knows that it's incomplete. And the major vendors of graph DBs support both SQL and SPARQL queries because most programmers prefer
to write SQL rather than SPARQL.
The only people who are confused are clueless academics who publish papers and dissertations that nobody who has a day job would read.
KI
> Q: What is the Semantic Web about?
> A: Webby structured data based on an entity relationship model that's
> endowed with machine readable entity relationship semantics.
Peter P. Chen published his paper on the E-R model in 1976. The basic notation is based on Bachman diagrams from the 1960s. The information expressed in E-R diagrams includes type constraints and cardinality constraints that are just
as important for formal mathematics as they are for practical DBs of any kind.
KI
> Ask the folks behind that [IBM Watson] project and they will tell you
> how much the LOD cloud contributed to Watson's smarts.
Yes, indeed. They gather up all that data from everywhere, and they store it and process it in -- guess what? -- DB2. A relational DBMS!
John
_________________________________________________________________
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