Nick @ Education.au

Patrick Lambe at Greenchameleon recently blogged a very positive reaction to a paper my colleague Sarah Hayman presented at the Developing and Improving Classification Schemes conference. His reaction was mostly to our concept of a Taxonomy directed Folksonomy, but he mentioned the idea of person-mediated serendipity. This is an area we agree is a vital concept, but we probably haven’t spoken about enough. (As an aside, this appears to be closely related to something Jon Udell refers to as “manufactured serendipity”, most recently in Data finds data, then people find people).
One key concept our team has been exploring is the trokia of resources, tags and people. The idea is that users should be able to use any one of those starting points to find useful information in the other two. For instance, if you are interested in the “education” area you can go to the education tag and find both resources related to that area and people who do work in that field.

Del.icio.us does this by displaying “active users” on a particular tag page. (eg, the education tag page). In one of our demos we took this concept a little further by displaying the active users in a tag cloud (which some are referring to as a guru cloud - although that makes me wince)

In our most recent demo project we used the edna resource collection to automate the discovery of authoritative del.icio.us users in areas we were interested in. Tom has posted the presentation we did on his blog, but the algorithm is worth outlining here (since I can hardly read my own whiteboard writing):

• For each URL in EdNA
• Check to see if it is bookmarked on Delicious
• If it is, save the usernames of each person who bookmarked it
• Monitor the delicious users who bookmarked the most EdNA urls

While that’s a very simple algorithm it has proved to be very effective. I did a presentation (see slides) which includes some hard data on just how well it has worked, but the best form of proof is usage: our edna information officers are already using its output as a source for potential inclusions into edna. That is despite the fact it was supposed to be only a demo system.

Something which isn’t discussed in the slides is the idea that this algorithm is tunable to either give more, lower quality or less, higher quality resources by changing the users we monitor. If we monitor only a few users who have a lot of resources in common with edna, then we tend to get a few new resources, most of which are suitable for inclusion in the edna resource database. By “dialing up” more users, with fewer resources in common with edna we get more resources, but proportionally less are suitable for inclusion.