Proposers of metadata standards say that
by reading a Web site’s metadata, an auto-
matic agent can understand a Web site’s
contents and simplify information retrieval.
Also, selecting the appropriate metadata
tags can give Web authors and publishers
some control over how their documents are
publicly indexed.
This edition of “Internet Services” dis-
cusses that, while the old META tag might
be alive and well on the Internet, the fledg-
ling Resource Description Framework
promises a richer description of Web docu-
ments.
The veteran
META tags, a very simple form of
metadata, have been around for a few
years. The 1994 HTML 3.2 standard
defined the META tag as an option that
could appear in the HEAD tag of HTML
documents. The syntax can associate an
arbitrary list of key-value pairs to the
document in which they are contained.
META tags frequently give a short
description and attribute a document’s
authorship (see Figure 1).
In Figure 1, the PROFILE attribute
points to a document that explains, in plain
English, the meaning of the tags used in
the META section. The last tag is used for
a digital signature, and the SCHEME
attribute provides some hints on how to
interpret its value. Although no standard
for META tag usage exists, nine of the 10
most prominent search engines use the
TITLE, Description, and Keywords tags to
index documents, and ignore all other tags.
Although generic search engines ignore
custom tags, domain-specific searches can
use them. In principle, a PROFILE can
indicate domain-specific tags, as Figure 2
demonstrates.
META tags work very well for limited
applications pertaining to a narrow domain,
but problems quickly arise if we want to use
this coding scheme to include a deeper
analysis of structured documents, knowl-
edge representation, and reasoning about
Web pages. For example, what if you want
to combine the previous two examples in a
hypothetical document in which someone is
writing a journal and is advertising some
artwork for sale? Although you can refer to
a list of more than one collection of META
tags in the PROFILE attribute, the same
property name (for example, Author and
Date) might be used under different profiles
with different meanings, creating confusion.
In addition, META tags only refer to the
document in which they appear, and you
might want to list properties pertaining to a
specific document section or entire collec-
tions of documents, directories, or Web
sites—without having to update each indi-
vidual document. Finally, we might want to
organize collections of META tag properties
into a hierarchy of classes. In an object-
oriented approach, general classes listing a
core collection of properties (for example,
20
IEEE INTELLIGENT SYSTEMS
The skinny on
metadata
Metadata surfaces as a hot topic in almost every discussion
concerning the future of World Wide Web standards. On the
Internet, metadata tags indicate a site’s indexing properties. The
tags automate the task of answering questions such as “What is
this page about?” and “Who is the author of this page?”
By Giovanni Flammia
Caliper Corp.
gflammia@alum.mit.edu
I
I
I N T E R N E T S E R V I C E S
Figure 1. An example of basic META tags.
<HEAD PROFILE=”http://www.w3.org/TR/REC-html40/struct/global.html#h-7.4.1”>
<TITLE=”A day in the life of someone”>
<META NAME=”Author” LANG=”eng”
VALUE=”Someone”>
<META NAME=”Rights”
VALUE=”My Company”>
<META NAME=”Description”
VALUE=”A Short Description of this document”>
<META NAME=”Keywords”
VALUE=”Work, Life, Dreams, Art”>
<META NAME=”Date”
VALUE=”Thursday, July 1, 1999”>
<META SCHEME=”MD5”
NAME=”Message Digest”
VALUE=”f96b697d7cb7938d525a2f31aaf161d0”>
</HEAD>
Figure 2. A PROFILE for an artwork Web page.
<HEAD PROFILE=”http://www.auctions.org/artwork-tags”>
<TITLE=”A document about artwork for sale”>
<META NAME=”Author”
VALUE=”Jackson Pollock”>
<META NAME=”Type”
VALUE=”Oil On Canvas”>
<META NAME=”Size”
VALUE=”100 cm x 300 cm”>
<META NAME=”Price”
VALUE=”$1,000,000”>
<META NAME=”Date”
VALUE=”August 1, 1968”>
</HEAD>
