Collaborative Cataloguing

Chris Dent

L505 Essay 8 2001-02-26

A word based system for describing the aboutness of pictures can be constructed by changing our model of where the act of representation takes place. Typically the rules for representation are established by some external agency - OCLC, technical services, Library of Congress, etc. What if we were to re-establish the point of representation activity as the patron group (O’Connor)?

This is a fabulous idea. As elaborated in his text, O’Connor’s idea is to allow the patrons of an image collection to both use and manage an image collection in a process of continuing refinement. He does not go into much detail on the actual implementation, so I’ll make something up here. Web based tools could be used for such a system. Imagine:

An organization gathers and digitizes a collection of images for a relatively known audience. These images are then published in a network-accessible database that allows several functions:

The images may be viewed by browsing through thumbnails that are sorted either randomly or by known categories such as date of publication, any obvious subject matter or artist (if any of these are known). Images of interest may be selected and viewed.[1]
Any selected image is displayed with the option to add “aboutness” commentary to the image record. Depending on the audience the commentary could be freely generated or be selected from a controlled vocabulary of some kind. If structured commentary similar to the Sara Shatford inspired method used in Cokolat’s Un Chien Andalou is desired a structured form would be used.
Existing commentary is available in both the browsing interface and an additional interface that allows hierarchical (if the commentary is hierarchical) and free-text searching.
Interested participants may view commentary written since the last time they visited and provide evaluative scores. These scores could be used to weight the commentary and/or allow other searches to adjust their queries for a level of “sureness”.

Obviously this kind of system requires a great deal of involvement from the audience. There are also considerable obstacles to overcome: any sufficiently large database of images with a sufficiently large audience could create a very large collection of text. At what point is there enough text? When is there too much?

(Programming is left as an exercise for the reader.)

Why is this system a good idea? There are at least two reasons: images are complex and at this time there is no effective way to search for images with images. Therefore, words must be used to describe the desired images. No one entity (person or organization) can possibly predict all the textual abstractions an entire population may choose. Images are not signifiers in the original, simple semiotic sense where a sign is the pairing, in a one-way relationship, of a signifier with a signified. Sure, the traditional demonstrations of signs are images (a picture of a horse with the word horse) but that image of the horse is very simple. Instead complex images are promiscuously significant, spawning off ideas galore in the mind of the viewer. Those ideas spawn off more ideas and more images [STC1] .

Searching by indicating the image of a horse would be exciting but few images of horses are alike. Additionally what do you gain? Most queries into an image database will involve both a subject and some descriptors (a sentence or at least a phrase instead of a word). What will you do if you want a horse that is jumping? Click on an image of a horse and then a representation of something jumping? That would be quite difficult to make a computer do unless one had access to an enormous, analyzed collection of horse images. Having such a collection is certainly possible, but improbable in the near future.

The alternative suggested above leaves us with smaller collections of images that have been described by many people [STC2] . Those many people hopefully have provided enough descriptors that make sense to the audience to allow the audience suitable access to the images.

REFERENCES

Cokolat, L. (1999). Providing Subject Access to Un Chien Andalou: a faceted classification approach.

http://skipper.gseis.ucla.edu/students/lcokolat/html/Lora_html/UnChien/ChienAndalou2.html

O'Connor, Brian. (1996). Pictures, Aboutness, and User-Generated Descriptors http://academic.emporia.edu/wcb/schools/SLIM/slim/boconnor/1/files/goodrum.htm

[1] An ideal system would maintain some state that allows images to be grouped, evaluated and then further regrouped without losing access to groups chosen thus far.