Comment Joy, Explanations, Further Directions (Score 2, Informative) 17
Internet Ninja, smart of you to point out the Google-Sets based visualization. I think that this dataset is potentially even more interesting then web-pages.
The only problem with Google-Sets, is that they were designed to be used with multiple search terms, not a single seed. Thus the data returned is often noisy, as can be seen by people questioning the links.
Wouldn't it be cool if Google's data were cleaned up enough that you could determine the exact degree of relatedness between any two concepts? That way, any area of interest could be mapped using a graphing tool like TouchGraph. And mapped in real time too, I am still getting around to generating a cronological record of how relationships between publications change over time.
My feeling is that this cleanup will happen, and visualizing the results will be the motivation for doing so. A text based list of similar items disguises the errors in the ranking because readers don't give that much importance to the order of appearance. A graph on the other hand can show hundereds of associations as opposed to 20 or so items that are usually shown in a list. Thus, a graph magnifies the errors in calculated similarity degrees, which makes it a powerful tool for improving the formulas involved in the calculations.
There is much more to be said about TouchGraph, but there will be more supporting material to say it later, so let me just finish with a couple of links.
Amazon has also released an API, and an TouchGraph powered Amazon Browser is in the works for that. Screen-shot
Work is being done to integrate the TGGoogleBrowser with the ODP and to represent hierarchical information using background colors.
More information about future devolompents is available via this thread on Webmaster World
Thanks, --Alex
