A while ago I came across Pivot, one of the experimental initiatives from Microsoft Live Labs. The idea behind Pivot is that it allows you to browse large sets of information in a visual way and allows you to find relations between information (at least that's my translation of it).
At the moment it is still quite static. You have to create a collection of items you want to view. Each data item should be categorised in multiple ways so you can look at it from different viewpoints (similar to what you do when using a pivot table to look at data from different viewpoints). And each item needs to have a visual representation because of the visual way the data is handled in the Pivot viewer. There is a helpful Excel sheet which allows you to define the collection and create the deep-zoom formatted images for the viewer... and that is pretty easy (as long as you keep you collection quite small).
I played around with a couple of pictures from New Zealand (where else ;-). They are all in two categories, region and distance from Wellington and I added some descriptions which show up in the browser when you zoom in.
It's rather cool to browse and filter through the collections like this. Especially being able to filter on one or more of the categories is quite useful when browsing one of the much larger sample collections. Unfortunately there is now way yet to collect items you found in your search.
The whole exercise got me wondering about two things, how do you come up with visual representations for non-visual information and how can you make it more dynamic. The example that allows you to browse wikipedia explores some ideas on the first question although I'm not sure on how static that collection is. And I am able to view my pivot browser history in the "pivot way" so that means that it must be possible to create the necessary xml on the fly... and yes... pages are converted to deep zoom images (and tucked away in C:\Users\<user>\AppData\Roaming\Microsoft\Pivot\Collection.dzc.Files)... this asks for further exploration because obviously being able to create a collection dynamically makes this far more interesting and usable.