Monday, December 18, 2006


I've spent way too much of my day exploring OCLC's FictionFinder prototype. Read more about the project.

The subject tag cloud that you encounter when first entering the system intrigues me -- the most commonly used subject appears to be "Marriage." I followed the subject "Missing children," vaguely thinking that I might encounter From the Mixed-up Files of Mrs. Basil E. Frankweiler. Nope. I searched for it and found that it's actually the subject "Runaway children." I wonder if I could have found it without knowing the title or that subject term? How do you know what subject term is the right one when browsing or searching? When is something under "Quakers" and when is it under "Society of Friends"?

I think that defaulting to genre for browsing is the right choice. It's a manageable length for browsing (at least for now), while the "subjects" list is quite long and "characters" is huge. See Thom Hickey's post on the difficulty in creating that character list. The awards list frustrated me momentarily -- I had to remember that the "Edgar" awards are actually the Mystery Writers of America awards and look under M.

The "settings" browse list results could be frustrating for some. I clicked on Mexico and found books where the subject was actually "New Mexico." It was great to see that books where the subject was "New Mexico -- Santa Fe" showed up under "New Mexico."

I searched on "voodoo," which is not the official subject terms (It's voodooism, if you care to know). I got books where voodooism is a subject. I got books where voodoo is in the title. And I got books where voodoo is in the description, such as "By the author of Voodoo, Ltd." I know it's a tough problem to index the assigned terms and other fields where relevant subject topics might be found.

The FRBRization display for a work is a sensible one. Who knew that Gaston Laroux's Fantome de l'Opera was available in Thai? Nice to see that I could follow the edition into WorldCat to see that Cornell owns it.

I did come across many examples of works that should have been identified as the same but were not for a single author -- H. P. Lovecraft. When I had the same experience in LibraryThing some months ago, I spent some time cleaning up the work relationships. I wonder why his works are difficult to identify and combine programmatically?

How do I combine browse types? I'm looking for books set in England that feature ghosts. There's an advanced search but no advanced browse. Maybe something like at Amazon, where one can narrow within facets: jewelery --> rings --> gold --> emerald.

I don't want this to sound like a rant, because it isn't. I think this is a really promising prototype, both as a FRBR experiment and as a subject browse environment. The fact that you can generate the browse lists at all is exciting.

No comments: