September 5, 2007
Tough Talk About Tagging
A few years ago, it seemed as if everyone was talking about folksonomies — Web projects that let users “tag” items with keywords and create their own collaborative categorization systems.
And to be sure, there have been plenty of folksonomic success stories. Sites like Flickr (which lets users post and tag images) and del.icio.us (which does the same for Web pages) have built up reasonably robust classification systems, and plenty of blogs — the Wired Campus included — now make at least some use of tags.
But some observers are beginning to wonder if the promise of tagging has gone unfulfilled. Despite the popularity of folksonomies, writes Matt Mower of Curiouser and Curiouser, “the state of the art in tagging seems firmly wedged in 2003:”
Tagging in 2007 seems to have advanced no further than a means by which one or more users of a site (or application) can group content around a loose framework of concepts. If you are lucky those concepts may express relationships but often they do not.
Along the same lines, Paolo Valdemarin says he’s still waiting for “truly advanced tagging tools” that will combine tags from different folksonomies and turn them into something useful. Professors who have experimented with Flickr and del.icio.us must surely have some thoughts on the topic: Have you gotten much out of tagging? Where should the field head next? (Thanks to David Weinberger for the links.) —Brock Read
Comments
Commenting is closed for this article.
Previous: On the Web, College Football Coaches Pull Out All the Stops
Next: House Votes to Narrow Technology Gap
I must say that the issue of tagging seems to be that there is no one term picked to represent a concept. The solution, in my opinion, is establishing a controlled language (as has been done with MeSH [medical subject heading] or Library of Congress subject headings) or creating a way to automatically map natural language terms so that people do not have to guess at which synonym is being used to stand for a concept.
— Amy Blevins Sep 5, 04:55 PM #
Applying some kind of cluster analysis or factor analysis-type of algorithm to delicious or flickr would probably result in interesting and I think useful results. Something like this is done with the Qunitura search engine (http://www.quintura.com), which I think works really well.
— Alan A Lew Sep 5, 05:36 PM #
Let’s make sure to define our terms clearly.
Tagging for personal or small group use is one thing. Can work well and be a massive time saver.
However, for the masses, it’s another story.
We’ve been cautious about tagging for quite some time on ResourceShelf. This post talks about name authority issues with names, pluralization issues and much more.
http://www.resourceshelf.com/2007/01/31/survey-american-tagging/
We also mention what Vivisimo, Clusty and others ar e up to in letting the content on the page speak for itself. With the metadata assigned by NLM, ClusterMed allows many ways to sort results dynamically.
True, this is not available for images/video but with Content-Based image retrieval coming on strong as well as speech search for video and audio, more options are becoming available. You can also add OCR search of images allowing you to search words inside an image. Finally, take a look at what Nexidia.com is up to with audio and video.
Next, time issues. Cataloging in the traditional sense takes time, a lot of effort and knowledge. Will everyone tag correctly and have the time to do it right.
Moving on, people trying to manipulate the system for personal gain. This is what happened in the early days of web search when webmasters were allowed to add meta-keyword tags to pages. It was abused so much, they are no longer being used.
If you read the post above as well as this one
http://www.resourceshelf.com/2006/03/26/a-few-reactions-to-newsweek-article/
you will see that we’ve been talking about these issues for some time.
MESH was mentioned in the first comment. Remember, that their are numerous controlled vocabularies that might be of use. Not the answer but potentially could help. We list a few of them here.
http://www.resourceshelf.com/2007/07/16/controlled-vocabularies-aod-thesaurus-the-alcohol-and-other-drug-thesaurus/
— gary price Sep 5, 06:21 PM #
Regarding Amy Blevins’ comment, I’ve heard that folksonomy-driven sites have been doing exactly this, only behind the scenes. I can’t remember the source any more.
— Kevin Hawkins Sep 6, 02:05 PM #
I would simply add that the community as a whole is badly in need of rigorous, well-designed experiments to test the relative costs and utility of alternative approaches to “tagging”. The underlying problem is – of course – that the enormous and expanding number of “digital information objects”does not seem easily
“tag-able” (or affordable) using conventional artisanal approaches. We are in very preliminary discussions at the Getty Research Institute about implementing such tests…
— Tom Moritz Sep 6, 04:45 PM #
One of the advantages of tagging is its simplicity. The same system with an easy interface can improve a lot.
At the moment “semantic tagging” with controlled vocabularies is in its infancy ( see http://jakoblog.de/2007/09/10/tagging-enriched-with-controlled-vocabularies/ ) and good interfaces for this are yet to be developed. Next step will be tagging with semi-controlled vocabularies like Wikipedia’s category system that can be changed easily. With tagging the whole history of information retrieval and subject indexing is reinvented but with better tools because traditional terminology failed to follow the developement of usable interfaces.
— Jakob Voss Sep 13, 11:56 AM #