I’ve tried, a bit, to popularize ethnoclassifications and freetagging in an essay just posted to the Adaptive Path site, Metadata for the Masses.
I have heaps more I want to say about it, and will over time. This essay was a challenge — trying to figure out how to talk taxonomies without getting mired in information architecture nonsense. I don’t know how successful I was. The goal was to introduce, and make relevant, the concept to people who might not be familiar with it. I also wanted to introduce some more notions to how to bring order to the chaos that is free tagging.
Feel free to discuss the essay in the comments area here!
I think your ‘paving’ allusion makes it make sense to a non-IA person such as myself [although, as a systems engineer, I do think about these things]. Control freaks [hi!] abhor self-tagging, but that sold me on it.
It’s an easy read and a good intro to something that is interesting and complex. Good stuff!
You should link your post back to some of the other discussions here – they were good.
(don’t mind me, I always tell other people how to write their posts 😉
I’m curious about how tools like Google Desktop will reduce or even obviate the need for tags and classification. As long as there is an index to search, I’m happy about 95% of time with any search I conduct. However, I suppose it would be nice to add tags to things I care about. In fact, I guess that I have started to do that with files I have been using recently. I add certain keywords that I know that Google Desktop will pick up, which in turn means that when I search these “free tags” make it easier to find.
(I’m just using Google Desktop as an example. Other similar tools — Longhorn search? — would probably work in the same way.)
To continue the discussion from WebWord, I don’t think the fuss over all the desktop search engines (be it GDS, Spotlight, Longhorn, Quicksilver, AppRocket etc.) is in any way changing the need for metadata. The reason metadata tagging on a community level exists for these things – links, pictures, musics and in the future when we have more bandwidth, videos – is because searches don’t do a good job of it right now.
Look at how bad Google Image search can be. It basically is almost completely reliant on the file name.
I wrote some more examples of community tagging as well as similar points to the above: http://www.ok-cancel.com/archives/post/2004/10/search_is_not_all_there.html
One thing I didn’t talk about was del.icio.us or link tagging. Link tagging is a bit more interesting because conceivably, Google -could- do something about those simply by accessing the contents of the bookmarks (or snapshot caches like Furl). While I believe bookmarks are unwieldly, couldn’t you just bookmark interesting things and have Google look through those as a means of constraint?
Good article and good comment above by vanderwal!
I’ve never thought the idea of a top-down Semantic Web could work, but I’m excited by the emerging folksonomy of Flickr and Delicious. I think we all need to have our own personal patchwork Semantic Web made by ourselves and those we trust. The problem with the current tagging systems is that they’re only one level deep. They succeed because they’re quick and require little active effort to create, but the lack of a heirarchical system causes confusion. Hopefully someone tagging “flow” would include other descriptive words that would combine to clarify meaning, but if “flow” were within a hierarchy it would solve the problem resolutely. But this complicates maintenance by a magnitude. How do you introduce hierarchy and still keep things simple? I hope someone can figure that out. Things like OPML can allow easy sharing of hierarchies. In the future I’d like to see taxonomies being shared like trading cards within communities, plugged in and out of master personal taxonomies based on our own personal preference rather than the dictates of someone at the top. Find someone who makes the subject of wine crystal clear? Plug their wine taxonomy into your master outline.
Tim — seems like the coincidence frequency of tags could be used as a measure of the strength of a link between to tags, which then lends itself to generating hierarchies using something akin to phylogenic trees. In any case, I think co-appearance of tags can get you pretty close to nested hierarchies.
Wouldn’t it make sense to build on the work of things like the Library of Congress Classification System? Librarians have spent decades creating metadata labels, with cross-references to synonyms.
The folksonomies/ethnoclassification systems do have associative appearance of tags, but they seem to be very rarely hierarchal in nature. The multiple tags are often descriptors of facets in natural language or vernacular. Hierarchies could be inferred, but rarely from the terms supplied. The multiple tags do imply possible synonyms as well as context.
The point is not to necessarily build a structured hierarchy, which many folks have problems using. The folksonomy is a more flexible system that has noise in the metadata, but is also has robust possibilities of enabling communicating across disciplines. The folksonomy is malleable enough to be capture emergent topics and frameworks in-synch with the pulse of change.
Reading Clay Shirky’s Many to Many (http://www.corante.com/many/archives/2004/08/25/folksonomy.php) will help provide some understanding of the robust nature of these tools.
Thanks for the links- I see now how a hierarchical system is sort of antithetical to what you’re talking about here. The Google vs. Yahoo example is a good one. But how do you make this portable? Is it trapped in the application you created it in, tied to everyone else’s classifications, or can it be transferred to other applications? It seems that all the standard data interchange formats these days are hierarchical. I’m not sure what the format of a folksonomy would look like. In what ways could Flickr integrate with Delicious?
It is somewhat of a Google vs. Yahoo shift, but it sits somewhere between with the adding of the metadata tags. The tags can be single or multiple additions to the object. The metadata augments the content of the object (if it is searchable such as text) or provides hooks for descriptions for those objects that do not yet have means of providing descriptions in themselves (objects that are media (video, photos, audio, etc.) do not have easy means of searching and grouping, see David Weinberger’s (http://www.hyperorg.com/blogger/) Wired article *Point. Shoot. Kiss It Good-Bye.* (http://www.wired.com/wired/archive/12.10/photo.html) for a better understanding of the problems associated with media objects.
Both Flickr and del.icio.us deliver tools that allows emergent categorization/folksonomy, but to slightly different ends. Flickr provides a simple means of tagging ones photos, which do not have their own easily discernible hooks to search on. The metadata tags that uses can apply give some description and context for the photos. This allows the person posting the photos to group them in a manner that serves them self well. The external benefit is that others can group photos with similar tags so to find an image they seek.
One of the things that the folks at Flickr are doing is working at providing synonyms to those viewing photos by their tags. Look at the *urban* tag (http://www.flickr.com/photos/tags/urban/) it also offers city, street, graffiti as possible tags of interest to the user. Flickr also serves up *See also* options, which have a relation to *urban*. This one means of leveraging other terms that are used with a similar relation to the user’s primary term, which may provide other options or even the exact image that is desired.
Del.icio.us provides a means to add metadata tags links, which similar to annotated bookmarks and can be grouped by the tags. The links and tags are also shared out and one can subscribe to the tags in their own *inbox* so to have items that are tagged with terms you understand delivered to you (there is some noise in this method as a term’s definition may change depending on the person applying the tag). Some of the noise can be mitigated by using tag pairs or other multiples to narrow down the links returned as well as define by associative context of the tags. Take the term mac, is it a raincoat, apple (fruit), Apple (computer), designer, road surface, or beauty product? By using a second associated term we could narrow down the type of mac we are intending on finding, by applying *ripe* we should narrow the results down to the apple (fruit), but this also narrows the items returned to what others tagged the link with or put in their link title or further description.
Del.icio.us is providing metadata tags to augment items with searchable text (titles and descriptions). The tags can be used for personal groupings or to provide a means for others with a similar vocabulary to find the items. In Google’s Gmail the metadata tagging is not shared, but intended for one’s own use to group and provide multiple structures for retrieving your e-mail.
How would Flickr and del.icio.us work together? Using similar terms one can find photos of an event (or object) and tie that to descriptions of the event. Take the recent O’Reilly Digital Foo event (http://digitalfoo.oreilly.com/), which has the following photos associated at Flickr (http://www.flickr.com/photos/tags/digifoo/) and the links at del.icio.us (http://del.icio.us/search/all?search=digifoo) and (http://del.icio.us/search/all?search=digitalfoo). The del.icio.us offerings are not as deep, but can get the idea of one of the possibilities of where things could head.
Today these tools and the means of providing social emergent tagging tools are still very young, but they are much better than what was out there a few years ago. Adding metadata to objects, be they text or media, is a chore, but the new tools ease this burden and are showing that a relatively flat metadata structure can be used to get to desired items.