search engine comparison
I hacked this quick and dirty vis in Sureshot cafe over the past few days. Some interesting speculation came from it: why is Live search and Yahoo’s, its acquisition target, already the most similar two search engines? Perhaps there’s been some motion behind the scenes for some time … (surprised look, music, … conspiracy)
but the core question is: does being “central”, or most similar to other search engines establish an engine as an authority? If I were to model this, I would start by assuming there is a set of correct documents for a search. Under this assumption, centrality is a second-hand measure of how well a search engine covers the correct document set, since other engines will tend towards the correct set. Of course other factors could conflate this interpretation.
It’s interesting that no two search engines are above 30% similar the datasets I’ve tried. Even searches top at abourt 60%, except for boundary cases where there are only a few results. I expected more overlap.
In terms of visual elements, I dropped a few of the more interesting touches because I felt the view was getting too complex. I’d like to explore them one of them here: Using occlusion to pop out the mean.
In this image, I wanted to the top right slice to stand out, because it’s greater than the mean. I wanted other large pieces to be visible, but not as prominent. A gradient from opaque to translucent, from the center, does a great job.
What’s most appealing to me about this technique is that it leverages our “pop out” effect, that the outer ring automatically draws our eyes and lets us quickly scan a list of these for outliers. The slices are arranged in principal directions for the same reason.