I was canoeing with my wife and infant son on the Snoqualmie river near Redmond years ago. That river rips, even in the summer and we were trying to get back to the launch site upstream. The swiftest current came rounding a bend in the river and we really had to dig in. I felt we were virtually flying and hollered out “We’re getting it.” but she corrected “No, we’re not.” I looked at the rocks on shore and saw she was right.
I kind of feel the same way about all the different tagging strategies that have evolved over time. There’s a bunch of them now and an imperfect evolution is something like SGML>HTML>XML>RSS>OWL. Periodically I have to dig in and learn a bunch of new stuff but there’s always some new version or standard just around the bend.
The problem addressed by all these is that humans interpret what they read by inference and computers can’t do that (or don’t do it well).
For us, those inferences are established by formatting, word choice, and personal experience. For computers, associations must be coded into the content explicitly. That sounds doable until you attack “personal experience”. I know there are approaches to that, but the blocking problem is easy to see in the real world: reasonable people differ.
So what am I getting at? Choosing a modest scope may yield more benefit than rigorous exercise. In other words, beach the canoe and portage across because there’s always more river ahead.
Applying this philosophy to tagging systems means using schemas (or the like) that are minimally or moderately complex. In my opinion, it is better to lose nuance than increase complexity.
OK, that’s an editorial. I promise next time to talk code.
Offer observations on the semantic web.