Joe Gregorio
You know what I want from Google? The ability to have concept grouping in the search results. Here’s how it would work: I search for a term like CVS. Now CVS has several different meanings, some of which are:
- It’s a drup store chain
- It’s short for Concurrent Versioning System
- It stands for Chorionic Villus Sampling, an early pregnancy test.
I want Google to detect that these results fall into different categories, and I want the results grouped by cluster with the groups highlighted by some mechanism, for example, different background colors. So all the drug store results are grouped together, all the versioning system results are grouped together, etc. That way I can quickly jump to the right cluster of results before scanning for a good link.
There’s a search engine at MIT called <a href=“http://www-math.mit.edu/cluster“>EigenCluster</a> (it used to be called Manjara and started life at Carnegie Mellon) that does this based on a complex data mining clustering algorithm. They have no editors, and sometimes the distinctions they draw are a little funny, but generally I find they provide interesting results.
It’s a meta search engine so the query performance is a little slow…
Posted by Tim Jarrett on 2003-06-18
Tim,
That’s very cool, and exactly what I was looking for. Now it doesn’t handle the CVS query very well, but it did a repectable job with boot:http://www-math.mit.edu/~drcheng/cgi-bin/main.cgi?query=boot
Separating out footware, boot disks, and boot camps. You can also click through to a neat little graphic:
http://www-math.mit.edu/~drcheng/cluster/visuals/boot.png
of how the clustering occurs.
Posted by Joe on 2003-06-18
Northernlight.com used to be my search engine of choice (before the went “private“) and had this functionality, it was so cool.
What they did was have blue folders on the left of the search results. When you searched you would get a list of the results alongside which was a series of folders with a few words next to them indicating a new level in the “hierarchy“. a search for CVS woul get you a folder with “Concurrent Versioning System” as the label for one of the folders, other fodlers might be labeled “Calgary Vocational Services” or “Curriculm Vitae“.
You could enter in a term and then drill down to the more specific results, maginc, hopefully it might be coming back soon http://northernlight.com/PressRelease.htm
Posted by Ben on 2003-06-18