[Top] [All Lists]

Re: [ontolog-forum] Lightweight, middleweight, and heavyweight semantics

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Tue, 06 Jul 2010 16:09:15 -0400
Message-id: <4C338D6B.2000908@xxxxxxxxxxx>
Bill and Mike,    (01)

WB> As David Eddy pointed out, almost all data has some kind
> of meaning (i.e., semantics).  What you're describing seems
> to me to be the degree to which metadata markup (e.g., tagging)
> is associated with a set of data.    (02)

Yes.  As I said, "Data with no explicit semantics of any kind."    (03)

And David added, "If there's no semantics, WHY has it been collected?"    (04)

That's a good point.  Nobody writes a program to process anything
that is not meaningful (except perhaps for a performance test).    (05)

MB> The property of meaning that we are interested in is surely "what
> meaning is intended?" rather than "what meaning is there?" i.e.
> explicitness. I don't think meaning is a property which can be had.    (06)

Yes.  I never use the word 'meaning' in a technical sense, since
different people have different purposes, feelings, and experiences
about anything.    (07)

MB> Then we can talk about explicitness of the semantics of some model or
> data resource, in terms of its relationship to some model theory, rather
> than as a thing in itself.    (08)

That's a good guideline.  But I'd be willing to allow any declarative
language that can be translated to some version of logic as a notation
for defining semantics.  That would include natural language that is
sufficiently precise to be translated to some version of logic.    (09)

But in many programs, the meaning is implicit.  For example, a file
of Comma Separated Values has no information about the semantics
of the data or why it was collected.  I would give CSV a zero on
the scale of explicit semantics (unless some metadata happened
to be stated somewhere outside the file).    (010)

WB> What you're describing seems to me to be the degree to which
 > metadata markup (e.g., tagging) is associated with a set of data.    (011)

Yes, the metadata is certainly one kind of explicit representation.
But when you get to semantic systems, the rules and constraints
blur the distinction between metadata and implementation.    (012)

WB> Where would data in a relational database fall?  Lightweight
 > semantics?    (013)

The rows in an RDB table are at the same level as RDF tuples
-- lightweight semantics.  And many applications (Linked Data
for example), don't go beyond that level.  In fact, some LOD
systems include CSV files -- but I hope that they would add
some minimal amount of metadata outside the file.    (014)

The constraints, views, queries, and triggers in an RDB can use
full first-order logic.  This level of semantics for an RDB is
roughly equivalent to OWL for the Semantic Web.  I would call
that middleweight semantics, since they don't state all the
semantics in detailed axioms.    (015)

The obvious example for heavyweight semantics is Cyc.  But
there are other AI systems that reach that level.    (016)

John    (017)

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    (018)

<Prev in Thread] Current Thread [Next in Thread>