Raw

Note to self and/or LazyWeb. Pluggd is an interesting-looking bit of Web 2.0ishness, from what it says at TechCrunch, it does speech recognition and searchability/granular addressing on podcasts. But like the toddler on Christmas day, I'm more interested in the box than its contents (there are plenty of libs around for speech processing and search). The use of a heatmap for relevance strikes me as a really neat idea.

 

It clearly lends itself to linear, continuous material like audio, but there's no reason a similar representation couldn't be done over largish quantities of text (how about a Sparkline version next to links?) and no reason it has to be linear. Imagine a search result that produced a 2- (3-, or 4-?) dimensional heatmap corresponding to relevance on a search on a whole site. Or one of those node & arc thingies, colour-coded. Mouseover to get an excerpt. Or what about a map of relevance correlation between two documents or sites? Say an examplar page on the y-axis, a page for comparison on the x-axis. Dunno. I like string.