Chris Harrison has developed a number of really interesting novel web and internet visualizations. I don’t have time to explore them in full but I thought them worth linking to.
Clusterball: pictured below is a visualization of Wikipedia’s category system. I only wish I could navigate links in this way. A very clear way of representing a small subset of a network as displaying its own logic. Perhaps this indicates that where you start (with a search term) will fundamentally change the network you navigate.
A very interesting visualization on this site is the graphing of internet domains according to the sites that a domain includes in the top 50 internet sites globally. Chris uses this visualization to image the increasing number of sites in ‘minor’ domains that are included in this ranking. The overwhelming increase in the dominance of minor domains in the ranks of popular sites is placed in stark contrast to the .au domain which has never had many sites in the top 50 but has been in a state of decline over recent years.
Chris’s Searchclock visualization offers a novel visualization of internet search traffic distributed over a 24 hour cycle. There are a number of other really interesting visualization and network projects on the site worthy of further exploration
The more I think about this, the more I think that tags are so important, but only because they provide flexible forms of remixing one’s engagement with a database – what we perhaps have not thought about yet is the way we might build a variety of flexible forms of data visualization into the interface as another way of “thinking through” things.
Tags do provide an unprecedented malleability to a dataset. They also present a whole bunch of problems (good problems). Look at any ‘power user’s’ del.icio.us tags. The tags for the most part are redundant – I mean their semantic definition is redundant. The most important development that tagging presents (at present) is the fact that a particular user has tagged something (as ‘anything’ at all). This value-adding isn’t simply via the endorsement but via the through-linkages that this endorsement ‘connects’. With this in mind I’d suggest that the key problem is working out how ‘flexible forms of data *sensualization*(?)’ might fold into a more ‘organic’ or ‘immediate’ form of tagging….and therefore a data-space with a greater degree of ‘communality’. By ‘communality’ I mean only a technical incipience for realizing those ‘through-linkages’.