Message-Id: <200011300356.UAA24450@dns.ccit.arizona.edu> Date: Wed, 29 Nov 2000 22:53:17 -0500 From: Thomas Heck <mailto:heck.3@OSU.EDU> Subject: Re: image description as metadata To: mailto:IMAGELIB@LISTSERV.ARIZONA.EDU
<pre>
<html>
At 09:27 AM 11/12/00 -0500, Kari Kraus wrote: <br>
<blockquote type=cite cite>I'm hoping someone on the list can point me to
a study or two on image <br>
description as metadata. My sense is that while controlled
vocabularies are <br>
a favorite talking point among information managers, the free-text <br>
descriptions that often accompany visual objects in a database haven't
been <br>
subjected to the same kind of scrutiny. </blockquote>. . .<br>
<blockquote type=cite cite>Any thoughts on how to define the relationship
between descriptive and more <br>
strictly classificatory categories of information? Is there a
theoretical <br>
body of literature in place that looks at the challenge of describing
images? <br>
A set of guidelines or recommendations?<b> </b></blockquote><br>
In my recent book, PICTURING PERFORMANCE: THE ICONOGRAPHY OF THE
PERFORMING ARTS IN CONCEPT AND PRACTICE (Univ. of Rochester Press, 1999),
I address this very issue as a practical retrieval problem. With no
preconception as to what the doctrinally "right" answer should
be, and even despite my formal background in librarianship (and
controlled vocabulary systems), my perception today is that the online
environment in which we live, with computers making "keyword"
searches the norm rather than the exception for most people, is making
free-text descriptions the preferred "ocean" in which to
"fish" n almost every situation.<br>
Why? (a) It's normative for the Web. (b) Natural-language
retrievals are already normative in most museum online databases. (c)
Keyword searches always seem to retrieve more "hits" in
databases with both controlled and free-text options than the controlled
field search option alone, even when you get the terminology right.
Experienced searchers do both types.<br>
A good example in the verbal world is the Dissertation
Abstracts database, which allows full-text retrieval of its abstracts
(since about 1980- ) as well as its titles. People, I have observed time
and again, seem to prefer to take their chances with keywords in this
database, rather than attempting to do controlled vocabulary subject
searches.<br>
A good example in the visual world would be searching the
ImageBase at the Fine Arts Museums of San Francisco
(<a href="http://www.thinker.org/" eudora="autourl">www.thinker.org</a>),
where the "keywords" search gets you into pictures via a search
of words in often quite lengthy verbal descriptions of each picture, done
behind the scenes, mostly by volunteers. Their words are randomized into
"word soup," and thence linked to the images to enable keyword
searches.<br>
Consistent? no... Professional? no... Effective? yes!
Productive of good results? yes! User-friendly? yes!<br>
If you would like more info on the book, point your browser
at
<a href="http://www.boydell.co.uk/PERF.HTM" eudora="autourl">www.boydell.co.uk/PERF.HTM</a>.<br>
Thanks for the interesting question!<br>
<div>___________________________________________________________</div>
<div>Thomas F. Heck, M.L.S., Ph.D., Emeritus Professor, The Ohio </div>
<div>State University Libraries ------- E-Mail:
mailto:heck.3@osu.edu</div>
<div>1221 Langston
Drive -----
Vox/fax...(614) 442-8226</div>
Columbus, OH 43220-3929 URL:
home.columbus.rr.com/heck
</html>
</pre>
This archive was generated by hypermail 2b29 : Wed Nov 29 2000 - 21:58:15 CST