This is the first of a set of raw notes from sessions I’m attending at SXSW 2006, presented with little editing.
Beyond Folksonomies: Knitting Tag Clouds for Grandma
On bottom-up, user-created taxonomies, both public (shared) and private — social bookmarking sites, dispensing with folder structures in favor of user-created / organic databases. Problems of cross-client integration, maintenance, etc. The wisdom of crowds: How to extract wisdom from a crowd? Is the crowd just the Borg, or is the crowd wiser than the sum of its members?
David Swedlow (opposablemind.typepad)
Mary Hodder (napsterization)
xxx Lawley (recovering librarian)
Site for panel: beyondfolksonomies.com/wordpress
A bottom-up way of organizing information. “On Intelligence” — invariance at various levels of meaning.
A taxonomy is a top-down / formal way to make sense out of things. Is Folksonomy just a fancy way to describe language itself, its inherently organic nature? No: it’s way to classify things – language and classification are not the same things. We don’t all get to decide what “cookie” is, but with folksonomies, we can classify cookies however we like.
Folksonomies are a sort of pidgin language.
There are a lot of UI questions that come up. Which engineering teams have just the right UI teams and know how to focus on the usability questions? What are the tools and feature sets that will get the usability all the way there?
Flickr and del.icio.us and Technorati are the three main folksonomy sites. Dabble will be another – will let you bookmark videos online and save them. There are, at last count, 44 social bookmarking sites.
My question: So where is the interoperability? My browser’s bookmarks should allow tagging, and should let me choose whether things go into the public or private cloud. My flickr and delicious tags should talk to one another. What do we need to make this happen? A common data storage format and DAV, or something like it.
When you change your tagging habits, you create your own mess. What’s the solution – rigid taxonomies? Then we’re back where we started.
Do people really use their del.icio.us bookmarks? Or do they just get excited about it, then end up using their traditional bookmarks after all. It does seem to be useful for discovering other people’s info.
The tagging process leads to “emergent semantics.”
Taxonomy vs. Folksonomy is a false distinction – you need both.
Critical mass is what’s needed for emergence to happen.
“Implicit tagging”: What’s the stuff you pay attention to most often? attentiontrust.org is building an attention tracker – an OPML file of all the stuff you look at it in a day, maybe with your identity associated, if you want that. The attention recorder will have an additional piece that will let you visualize how you spend your time online.
Online video/audio is a critical place to tag, since it’s not inherently searchable.
The value of tagging is that a human has done it, not a machine. Automated tagging will lose all the context, and the quality of information will quickly degrade.
Implicit vs. automatic tagging: Not to take work away from humans, but to make use of the work that’s already been done by other humans. Trusting the crowd: We do this with the popular Bayesian spam filters, such as DCC/Razor, or PopFile. The crowd has wide agreement about what constitutes spam.
To go “beyond” folksonomies: Tools need to integrate tagging interfaces. Windows Vista will included a lot of tagging features. OS X already does… but does anyone use them? You have to proactively tag most content with metadata.
Librarians tend not to believe in the “wisdom of crowds” as much as others do. It can be useful, but for most things, do we really want the wisdom of crowds? The popular vote doesn’t always result in the best results. And do we want broad, vague groupings like “Friends?” I don’t have “just a bunch of friends” – there are lots of people I trust whom I don’t count as friends, etc.
People already know how to do folksonomies – we’ve been creating folder structures forever… BUT folders are hierarchical, while tags are heterarchical, so it’s very different. And folders aren’t social, of course.
iTunes has a taxonomy but it’s influenced by money and politics, which is wrong.
What sites are doing it well, and what are the best practices?
itags.net offers the spec for doing distributed tagging with CC licensing bound to the media object. With rich media, link degradation is really high.
Odeo had to remove their tagging service because its Ajax interface would glow when an object was tagged; people liked watching it so much that they’d tag and then delete garbage just to watch the glow, thus polluting the database.