I’m re-posting this here to force me to deal with it more productively. Jon Udell totally nailed what’s wrong with deeply structured taxonomies, and the different bags-of-keywords approach successfully employed by Flickr and del.icio.us.
What needs to be done is to show how the bags-of-keywords approach could be employed in such a way as to allow the librarians of the world to relax (i.e., can it provide an efficient way of navigating a large-scale repository or repositories, or is it effectively limited to small scale - personal or workgroup level repositories? Gut says it could scale, but there isn’t enough data to get the library geeks to chill.)
Did a quick skim, and he’s referencing the book “E-Learning in the 21st Century: A framework for research and practice” that was put together by Randy Garrison and Terry Anderson. Randy is the Director of The Learning Commons, while Terry is up at Athabasca University, and was involved with the development of CAREO.
I disagree with dcannell’s “thing of the past” verdict, in that I strongly think that learning object repositories are valuable (I’ve built a couple, and am working on something new), but that their role must change drastically if their value is to be realized.
OmniWeb 5.1 Beta 1 is available - includes updated WebCore (YAY!), and a bunch of other stuff… Now the text shadows render properly in css, and gmail works, and a bunch of other stuff. Thank you Omnigroup!
Well, not exactly alive, but it’s working. It’s freaking working! I can now insert new records (creating new xml documents) and save them, run queries, edit records, and save those changes. Whew.
There is some final cleanup I want to do (finish migrating from DOM4J to JDOM, clean out all of the debugging spaghetti outputs, and comment the heck out of it), then I’ll package it up and drop it on Sourceforge.
Just came across a link to the new Snap search engine. Not sure if it’s doing anything drastically different under the covers, but they are trying some new stuff in the UI (which may be a bit densely packed, actually).
The biggest thing they are trying is a javascript live-updating results display panel. You do your search, then refine it in another field, with the filtered results displaying in almost real time. No visible roundtrip to the server (but there may/must be one in there somewhere…)
It just struck me that the shortcut I described to allow me to easily search del.icio.us and Flickr would have been darned near impossible if they’d used some form of complex nested taxonomies instead of employing bags-of-keywords.
The more I play with tools that use bags-of-keywords, the more I see the power of them. Sure, you lose some of the elegance on the metadata side of things (hard to define relationships and parent-child links with bags of keywords, easy with taxonomies), but do users really care (or understand) that stuff anyway? Just enter a few more keywords, and eventually you’ll have described it as richly as you would have with a Library of Congress classification, without having to memorize LoC…
Just added another really handy shortcut to OmniWeb 5. With it, I can just hit “command+L” to get to the address bar, and enter “del xml database” and del.icio.us will spit out all bookmarks matching the words (after del, of course - in this case I’d get all bookmarks tagged with XML database)
So insanely simple, yet amazingly useful. I have similar shortcuts set up to hit the Wikipedia, and some of the mailing list archives that I use, oh yeah, and Flickr tags, too! Pretty handy stuff. Thanks Omni! (I’m sure this stuff could be trivially adapted for use in Firefox or somesuch, too…)
The Laszlo Platform, which is kinda like Macromedia Flex, was just released as open source (under the Common Public License).
I’d looked at Laszlo Platform a while back, and it was pretty cool, offering dashboard-like stuff, sorta like Macromedia Central meets Sherlock meets Konfabulator. Kinda cool for dynamically generated flash content and widgets.
Anyway, in the article, Cory attempts to debunk the myth of metadata. He came up with 7 problems (the headings are, I believe, intentionally exaggerated and provocative):
People lie
People are lazy
People are stupid
Mission: Impossible – know thyself
Schemas aren’t neutral
Metrics influence results
There’s more than one way to describe something
I think the points may be quite valid (if a bit overgeneralized) when applied to “user entered” metadata. If Uncle Billy is adding something to APOLLO, he’s not likely to be a reliable/consistent/complete metadata tagger. However, if a librarian is tasked with cataloging a set of resources, these rules (hopefully) don’t apply, and the structured metadata becomes trusted and valuable.
Well, breakthrough may be a bit much, but we did figure out some stuff today, and it’s going to work quite well…
King just spent an hour with me pouring over the JavaXStreamDBAdaptor code, trying to help me figure wtf was going on (i.e., it wasn’t behaving as expected, leaving it unable to properly insert new documents).
We were looking at the code, and then King had an idea. “Open the EOModel for the xml database,” he said. I cracked it open. “Try it with the _gid attribute set a a non-class-property.” OK. I hit the little diamond widget to tell EOF not to let me access the _gid primary key attribute directly. Recompiled, tweaked some code, and BOOM. It was kinda sorta working. Almost.
