Vivisimo - clustering search results

Vivisimo is a search meta engine, whose main feature is the ability to cluster the result pages, based on content. This is especially useful when the search word has two or more distinct meanings. One such word is tank, meaning either a (usually large) receptacle or armoured battle tank.

Searching Vivisimo for tank gave 198 hits. Most, if not all, really belong in two main clusters, receptacle and armoured battle tank. These two clusters are not detected by Vivisimo, only their smaller sub-clusters. Still, this is a very impressive performance by a machine. (The movie Tank Girl may be about both water tanks and battle tanks, if so, it belongs in both main clusters. I can't tell from the web page, so we have to wait for a new release of Vivisimo - one that watches movies.)

Such clustering would be facilitated by metadata in (or outside of) the web pages. Suitable Dublin Core DC:subject URIs for these clusters could be


Webmasters may be less than truthful in their meta markup, so analysis of content and incoming links will stay important.


Skriv en ny kommentar:

Husk meg ?




Trackback-URL for dette innlegget: