After lunch, and onto breakout sessions. I've headed upstairs to the "emerging tech" track - we have David Weinberger and Paul Gillin talking about tagging taxonomies. This is a smaller session (I expect that most of the crowd at this show is in one of the PR oriented tracks). David is writing a book - "Everything is Miscellaneous" (due out in 2007).

The premise of the book is that everything is going online, and we then have an opportunity to create folksonomies via tagging. In particular, he wants to compare and contrast with the pre-existing, traditional management of knowledge. The traditional method is a lot like sorting laundry - if you have a book, it goes in one section of the bookstore (etc). This is a limitation on how things can/should be categorized. So what do more wide ranging taxonimies give us? That's what the book is about.

So to start: "What is a tag?" - tags are, of course, metadata attributes in ordinary (user defined) vocabulary. The important usage is not author defined, but reader defined tags.

Paul: We've had metatags in HTML for a long time now - what's changed? Applications that make use of them are starting to appear, and support user defined tagging. In other words, communities that care can create their own taxonomies. Things like del.icio.us and flickr are examples. Where it helps is letting us escape from the folder system when confronted with tons of data (eg: digital photos). What about trying to tag tons of data - David agrees that we need automated support from our tools (all the way out to the camera - for instance, combine GPS with a camera, and you can get good automated support). Add in things like facial recognition. The bottom line is, you need automation support.

Another form of automation: post photos from a social event like this conference, and invite people to tag them (distributes the task across time and space).

What about the ambiguity of tags? Generally speaking, we aren't looking for completeness. It doesn't work for people (eg, lawyers looking up case law) who need complete results, but it does work for most of us. For instance, searching for pictures of "London" just to get an idea as to what to visit - if you miss some because they were tagged "Picadilly", it doesn't matter that much.

Paul: Do you see bleed over back to the physical world? Already happening (RFID, newspapers, package bar code lookup, etc). We are already seeing the editorial control function moving to the web - Digg, Reddit, etc. - users are determining which stories matter to the reader community. What about getting more ability to read what other people are reading (OPML reading lists came up here).

Interesting point from David - we need metadata for the tags themselves: where and when they happened, public/private, language used, etc, etc. That part should be automatable. A question about search vs. tags: Troy blogged about that this morning :)

Very interesting point brought up by Paul - your public image is not (and has not been for awhile) under your control. Other people can tag you any way they want. Along those lines - what about suggested tags? For instance, tags to use for a conference, for posts about a natural disaster/event (etc, etc). Sure, it's not bottom up, but it could provide useful guidance.