Gert Says
a rather irreverent discusssion
Drill Down
In Web Schemas we often talk about drilling down to the desired information. I propose that we might consider searching for images in this same vein.
Drilling down refers to Web searches which are refined by a hierarchical arrangement of relationships which is often shown like this - Product >Software> Image management. Of course you can also drill down in a traditional cataloging system which organizes information in categories such as botanical phyla. The difference here is a flexible structure which permits many relationships. It is done with Keywords, an inexact way of describing an image but one which may permit greater flexibility in searching.
In the inexact column, keywords are usually not organized according to relationships or hierarchies. For example the word interior can have many meanings depending on its context - interior design, interior of a palace, interiors of the mind, going to the Interior. Also while there is a controlled approach to developing keywords in which there is a prescribed list of words which describes an image or the important facts describing the image in a way one hopes satisfies a searcher's need, there is also an approach where every word used to describe an image in say a notes field is searched as a possible keyword. In that situation the user must guess as to what words are being used. Some people attempt to organize keywords to make them more exact by indexing or establishing hierarchies, such as in Canto's Cumulus ( an media asset management software)
These relationships can inform the user when they are headed in the wrong direction. But what if it was the truly wild west and they could create there own rules of organization for each search. They would have available all the words that have been used to describe an image either in a descriptive field or long list of key words and they could organize their search according to hierarchies ( use the | pipe| command to do so) which they created. Say one user in an architects office is interested in examples of precast high-rise buildings in frigid climates she might enter - precast | high-rise | intemperate, but another is preparing a proposal for a 100,000 sf, medical office building in Washington State and needs some images of comparable work. He might look for office buildings | medical | mid size | NW USA. The additional benefit is, if there is no example at the more restricted search level, they can just go back up one.
Now some of you might say a user can still do that in a well worked categorical system and they can. Actually I am not saying the unformed structure is inherently valuable, but rather that a more flexible approach to accessing the information is. Also the work effort to create such a structure for a unique image library whose sole purpose is retrieving images not storing information about what the image represents, should be evaluated. The only reason I can see organizing descriptive text in that situation is it is the easier to describe and verify.
There is however, some text information which is particular to each image but not usually used as descriptive text for search purposes, such as storage location. Some other examples of unique data might be creator, publishing rights, file size, file type. This data is best stored in a unique field which can be verified and is necessary to use of the image, though not to find it.
For that, think about letting the user to plot how they will look for images, rather then you trying to figure out you think they will. You can do this by creating search screens, which encourage and help your user to organize their searches.
1/30/00
