Tuesday, July 08, 2008

Open Shelves Classification

LibraryThing is building the Open Shelves Classification (OSC), a free, "humble," modern, open-source, crowd-sourced replacement for the Dewey Decimal System.
The vision. The Open Shelves Classification should be:
  • Free. Free both to use and to change, with all schedules and assignments in the public domain and easily accessible in bulk format. Nothing other than common consent will keep the project at LibraryThing. Indeed, success may well entail it leaving the site entirely.
  • Modern. The OSC should map to current mental models--knowing these will eventually change, but learning from the ways other systems have and haven't grown, and hoping to remain useful for some decades, at least.
  • Humble. No system--and least of all a two-dimensional shelf order--can get at "reality." The goal should be to create a something limited and humble--a "pretty good" system, a "mostly obvious" system, even a "better than the rest" system--that allows library patrons to browse a collection physically and with enjoyment.
  • Collaboratively written. The OSC itself should be written socially--slowly, with great care and testing--but socially. (I imagine doing this on the LibraryThing Wiki.)
  • Collaboriately assigned. As each level of OSC is proposed and ratified, members will be invited to catalog LibraryThing's books according to it. (I imagine using LibraryThing's fielded bibliographic wiki, Common Knowledge.)
I also favor:
  • Progressive development. I see members writing it "level-by-level" (DDC's classes, divisions, etc.), in a process of discussion, schedule proposals, adoption of a tenative schedule, collaborative assignemnt of a large number of books, statistical testing, more discussion, revision and "solidification."
  • Public-library focus. LibraryThing members are not predominantly academics, and academic collections, being larger, are less likely to change to a new system. Also, academic collections mostly use the Library of Congress System, which is already in the public domain.
  • Statistical testing. To my knowledge, no classification system has ever been tested statistically as it was built. Yet there are various interesting ways of doing just that. For example, it would be good to see how a proposed shelf-order matches up against other systems, like DDC, LCC, LCSH and tagging. If a statistical cluster in one of these systems ends up dispersed in OSC, why?

No comments: