[Fwd: connection to natural language in schema defs [was: Comments to MCF XML proposal]]

Attached is a message I sent way back...
It's still relevant. I think we're setting a bad precedent
for the use of RDFS:Comment.
In stead of:
=======
http://www.w3.org/TR/1998/WD-rdf-schema-19980814/#dublincore
<rdfs:comment>The name given to the resource, usually by the Creator
 or Publisher.</rdfs:comment>
=======
Please write:
--------
<rdfs:comment>For Thing--title-->String read: Thing is
 entitled String.
</rdfs:comment>
--------
Hmm... maybe I'm suggesting a different facility, which
is some stuff to facilitate machine translation of
schemas to natural language. Consider:
<nl:explain xml:lang="en" nl:about="#Title">
 The title of <nl:propObj> is <nl:propValue>.
</nl:explain>
-- 
Dan Connolly
http://www.w3.org/People/Connolly/

Forwarded message 1

  • From: Dan Connolly <connolly@w3.org>
  • Date: 1997年7月03日 12:25:18 -0500
  • Subject: connection to natural language in schema defs [was: Comments to MCF XML proposal]
  • To: guha@netscape.com
  • CC: Ora Lassila <lassila@w3.org>, w3c-dsig-collect@w3.org, w3c-labels-wg@w3.org
  • Message-ID: <33BBE07E.49B2@w3.org>
Ramanathan Guha wrote:
> (ii) property value inheritance is used to specify things like "all dogs
> have
> 4 legs".
That snippet explains the whole concept, both to the technical
and the lay community, it seems to me.
I think it's vital that the metadata specs make explicit the
connection between metadata and natural language. Consider
the application to law, for example.
I want the metadata architecture specs to set a precedent so
that the schema definitions themselves will make the connection
to natural language as you did above.
In stead of (or in addition to) schema definition tables
like:
============
http://www.w3.org/TR/NOTE-MCF-XML-970624/#secA.2.1
A.2.2.1 Authorship Related property types
authorIndividual 
 The individual person(s) who is(are) the authors of the content
object. The entries are not names
 of the authors but references to objects corresponding to the
authors. The name, email address,
 etc. of the author can be specified on that object. 
authorOrganization 
 The organization which is the author of the content object. 
author 
 The generalization of the previous 2 property typess. The is a
superPropertyType of both. of them. 
editor 
 The agent that is the editor of the content object. 
publisher 
 The agent that is the publisher of the content object. 
contactAgent 
 The agent who is the "contact" for that piece of content. Typically
the person behind
 "webmaster@xyz.com". 
copyright 
 The copyright declarations. The range is page addressing the
copyright and other legal issues. 
============
I would really like to see each property (or set of related
properties) used in a natural language sentence; for example:
contactAgent:
	for contactAgent(content*, agent*), read:
	"The agent to contact regarding content* is agent*"
Descriptions like the one above too often have implicit
parameters. For example, in one of the "collections" drafts a while
back (can't find it now) there was a "print" property
whose description was something like:
	print
		Boolean. whether to print this document or not
A naive reader might build a model where print is a boolean
functional property of a document. But it's not! Document X
might be printed in the context of one collection, but not
printed in the context of another. The property is a relationship
between the collection and the member in question. So
the description should read:
	print:
		for print(collection*, element*) read:
		"When collection* is printed, element* is to
		be included in the print job."
-- 
Dan Connolly
http://www.w3.org/People/Connolly/

Received on Tuesday, 18 August 1998 10:56:30 UTC

AltStyle によって変換されたページ (->オリジナル) /