Unit 6/7 - Cataloging Woes

One of the issues I was disturbed by in cataloging my collection using EPrints was the ability to select more than one Subject Heading per record. On the surface, this seems really great because it's inherently flexible. However, since every cataloger is different (especially self-archivers), I see the potential for major disaster when browsing by subject in this type of environment. Compounding this problem, at least in my collection, is the inability to exploit full-text searching in multimedia collections. Unless each video/audio file is transcribed, the metadata needs to be extremely descriptive.

While the Subject Headings are useful in a general sense, they tend to be quite broad and would only begin to filter results within a very large collection. On the other hand, tags have the capability to be SO specific that they might weed out too much in a browsing environment. The potential for inconsistency in self-archiving is so great, but I found that Drupal's tools for creating a more guided entry process with customizable non/required selection and entry fields could eliminate at least some of these problems. I keep straddling the fence between micromanagement of these elements and Shirky's points regarding the irrelevance of such efforts.

I have been approaching cataloging and metadata creation inconsistently on purpose. Though I try to retain some amount of uniformity in each of the records, I have also chosen to add some keywords/tags to records and omit them in others to really see how this affects the user experience in both browsing and searching. This behavior has produce rather predictable results - even I can't remember which tags are linked with certain objects and I have to do some digging to get back to the object/s I was looking for.

On the other hand, I review user queries and results on a daily basis at work where we strive to create catalog records that are consistent as possible without any user input. Day after day I see skewed results and irrelevant returns resulting from pretty simple searches in a highly-controlled vocabulary due to idiomatic expression variations, plural terms, compound keywords, etc. So, I guess I still have hope for a system that uses both controlled vocabulary and free-tagging.