Computerworld

Grouping Web sites together

Have you ever wondered which Web sites are most similarto your own site? Now you can see a graphical diagramof your site surrounded by these others, thanks to anew online graphing tool by TouchGraph.com.

TouchGraph piggybacks upon the "related" feature of theGoogle.com search engine. Google doesn't classifyanother site as related to yours simply because yoursites link to each other -- it uses its ownproprietary rating scheme to group sites together.

You can easily see a linear, text listing of sitesrelated to your site -- or any URL -- by searching forthe site name at Google, then clicking the "SimilarPages" link that usually appears below each listing.

(This works best with well-known sites. Small andpersonal sites often have no "related" data at Google.)But TouchGraph CTO Alex Shapiro argues that displayingrelated sites as he does, in a two-dimensional orthree-dimensional graph, reveals relationships thataren't at all clear from merely examining a linear list.

For example, TouchGraph's home page links to a graph ofsites similar to the National Science Foundation. Theimage reveals a web (no pun intended) of NSF'srelatedness to various government agencies,foundations, and the like. More important, the diagramindicates sites that are closely related to each otheras "clusters" within the graph.

As part of a discussion of his concepts atWebmasterWorld.com, Shapiro says images such asTouchGraph's may initially show all the connectionsbetween 30 different sites, representing 900 possiblerelationships. This is far more detail than would besuggested by Google's typical list of 10 similar sites.

As a Java application exploiting Google's developerAPI, TouchGraph suffers from the limitations of thesearch engine. Sites that don't rank highly at Googletend not to have rich information on their ownrelatedness. (Observers have also noted that Googleseems to have developed its Similar Pages technologyin part to catch and penalize bogus sites that set uprings with hundreds of links to each other.)This is a limitation that Shapiro is well aware of. Histechnology, however, has usefulness far beyond searchengine data. He's already, for instance, participatedin experiments to map Amazon.com's "buyers of thisbook also bought" relationships. TouchGraph is aninteresting tool that's worth a look to see what youmight find.