Tagging’s Long Tail

Tagging systems offer a fascinating opportunity to study how people tag and what collective wisdom can be generated from the masses. Tim Spalding, in his a recent post at Thingology, The Long Tail of Ann Coulter, observes that tag use in LibraryThing resembles the Long Tail principle. That is, a few tags are used a great deal to describe a given item, while other tags are used just once. These singleton tags reflect the idiosyncratic nature of individual taggers.
I’ve been thinking about the value of these singleton tags, without conclusive results, in connection with MTagger, the tagging application we built at the University of Michigan library. With the 8.5 million items in the library catalog, or even the 55,000 web pages on our site, will enough tags ever be applied to enough items to make it a useful mode for a newcomer to find an individual item, or are they just an aide-memoire for the person who applied them? In other words, do tags way off in the Long Tail matter?
The more I’ve pondered this, the more I realize that it’s not an either-or question. Tagging, at least in the library environment, is most valuable as a personal collection tool. It offers a way for library users to bring together things that seem similar to them for their own purposes. The real value of tagging is like that of a library: it’s the collections, the constructed universe of things that someone (a librarian, a subject expert, a user) brought together. While my tags may prove of no value to anyone else in finding a particular item, the mass of items I’ve used that idiosyncratic tag on may very well guide a future user in resource discovery.

Tagging and Taggers

A recent research paper, “Can Social Bookmarking Improve Web Search?” by Paul Heymann, Georgia Koutrika, and Hector Garcia-Molina, draws numerous interesting conclusions about the effect of taggers and tagging on findability. The authors used del.icio.us as the source for tags.
Several of the results they found:

  • “Tags are present in the page text of 50% of the pages they annotate and in the titles of 16% of the pages they annotate” (p. 8). It seems that taggers are not particularly original in their tagging.
  • “Pages posted to del.icio.us are often recently modified” (p. 4) and “approximately 25% of URLs posted by users [of del.icio.us] are new, unindexed pages” (p. 5). By monitoring tags of interest to you, you can find out what’s new more effectively than you can by setting up standard search queries.

Their closing section, in which they discuss how tagging could be improved in the long run, bears quoting at length:

In terms of tags, we believe that user interface features could have a large impact on improving the quality of tags for search. For instance, interfaces that recommended tags not in the page, or not common for the given domain, might help alleviate those two problems. Another approach might be to have domain-specific sites (e.g., photography) which might have higher quality tags due to the shared context of the users.

New Tagging Tool at University of Michigan Library

I’d like to talk about a tagging project we just launched at my workplace. MTagger is a social bookmarking tool that we’ve integrated into several University of Michigan library resources. A tag cloud now appears:

Like del.icio.us and many other social bookmarking tools available on the Internet, MTagger allows users to bookmark and tag web pages using language that makes sense to them. Anyone can see tag clouds on pages and search MTagger; only users with valid U-M network logins can apply tags. (Individuals can, of course, opt out of sharing their tags with others if they choose.)
Unlike these other tools, MTagger offers the concept of “Collections” — letting users restrict their searches for similarly tagged items to a specific collection (library catalog records, images, web pages, etc.). While tags themselves would allow people to serendipitously find items in other collections, the “Collections” metaphor will, we expect, help drive home that the library offers more than books, electronic journals, and databases.
More important than the tagging functionality itself is what MTagger will allow our faculty, staff, and students to do. MTagger brings a social component to research that we have not previously had. It will allow users to share knowledge about library resources with each other, to enable quick-and-dirty subject guides to be produced, and — we hope — to bring researchers together via their individual tag clouds. As research moves online, chance meetings in the stacks of researchers with overlapping interests become even more rare. Through tagging, we hope to be able to recreate some of those synergistic interactions as one researcher finds a tag of interest, and through that, the other researcher.
Oh, and just to keep this in the realm of libraries and RSS, anything that can be searched within MTagger can be accessed via an RSS feed.