structured metadata using A

Bruno Kestemont (bkest@ulb.ac.be)
Mon, 12 May 97 10:41:07 +0200


Date: Mon, 12 May 97 10:41:07 +0200
Message-Id: <9705120841.AA06960@is1e.vub.ac.be>
To: www-html@w3.org
From: bkest@ulb.ac.be (Bruno Kestemont)
Subject: structured metadata using A

This is a complement for Stu Weibel's message (see bellow: HTML metadata for 
other web objects), and a proposition for html.

Excite (www.excite.com) argues that meta tags written by the autor of a page 
could easilly lie (e.g. for commercial attraction).

Therefore, I would suggest a 5th method:

using <A> tag in order to give a meta-name to any element provided on the 
screen.

example:
<body>
Welcome to 
<a name="keyword">sustainable development</a>
home page.
The author of this page is
<a name="author">My Name</a>
</body>

The names of the entities would follow html conventions (Dublin core, ...). 

Advantages:
-no necessary duplication of certain metadata (author, description, ...);
-more reliable (more difficult to lie);
-can drive the search engines to the more representative part of the document;
Disadvantages:
-more difficult to deal with standard terms or non textual objects (but meta 
tags in the header can be complementary).
-the "content" part of the meta information must be somewhere on the screen, 
at least in very small case!

To go further on this idea, Aurelien Slodzian (VUB) added the possibility 
even to describe a metadata structure of a document using the META tag for 
the structure and the A tag for the contents. This idea makes the system 
mutch more generic. Its "Agent Markup Language" aims to allow software 
agents to identify structured information within a html document sent to the 
public with a nice display. This appens often when a query to a web database 
results in an answer to the client with no apparent structure. The idea of 
AML is

"1.to place invisible marks in the HTML page, identify the text which form 
the textual representation of the information ; 
2.to place in an other page, or in the header in the HTML page a model of 
the data structures. "

I invite you to read the (draft) paper on 
http://artipc5.vub.ac.be/bloody/aml/aml.htm
and to comment to it.

Bruno Kestemont

Stu Weibel wrote:
>
>As I have said earlier, the PICS 1.1 label facility is not of interest
>to the metadata community.
>
>The PICS-ng specification (which is not, as far as I know, now
>available publically, in that it is being discussed by the working
>group) will allow strings, substructure, qualifiers, and repeatable
>elements.
>
>Further, there are 4 association models of PICS metadata and resources:
>
>   1. meta embedded in the resource header.  
>
>      easy, requires no additional infrastructure, and readily
>      harvestable, but perhaps not the best model for efficiency
>      (bloated headers) and maintainability (keeping metadata
>      consistent as resources are mirrored or copied will be a problem).
>
>   2. resource embedded in a metadata wrapper.
>
>      a useful model for wrapping images, for example.  same advantages 
>      and disadvantages as #1
>
>   3. coupled metadata
>
>      metadata is tightly coupled to the resource, may travel with it, but
>      is not embedded and can be handled as an object in its own right
>
>   4. third party metadata
>
>      metadata is linked to the resource by reference only; may or may
>      not be created, maintained, distributed by the manager of the
>      resource itself.  requires additional infrastructure (eg.
>      database management tools)
>
>      this is the model that third party rating providers (label
>      bureaus) will be based on, but it is also the model for
>      distributed resource cataloging that already predominates in the
>      library world.
>
>
>  Models 2,3, and 4 could be applied to non-html objects
>
>stu
>
>
>      
>
>