(and some wisdom on implementing "Bottom of the Page™")
MLT (More-Like-This) features are not so easy to implement if they are conceived as afterthought to creation of content. In our content management system (CMS), we use keywords to associate and group related content and pictures. But there is much more to MLT than simple keyword based associations.
This week, Hiryoung and I have started a new initiative that we call eXtra. The project is to develop associative technology and tools that will suggest products and links based not on human intelligence, but based in acquired heuristics.
Some of the parameters to consider:
- Aliases -- The software must learn Ganapati and Vinayaka are same.
- Page Rank™ing -- the more the page is linked, it is more relevant, the more a page or product is referenced, it is more relevant
- Physical proximity -- this is a very important parameter, building on human intelligence. We can safely say all the files in a folder (or products on one shelf at a store) are related to each other in some way.
As always, the commercial interests would collide with the academic interests of this effort. So we'd have to consider ranking a sponsor higher than the computed value of his link.
On some pages at Kamat's Potpourri (example), if you scroll down (hence the name Bottom of the Page), you will see a prototype of this experiment. I am surprised at the relevancy already!
See Also:
• The Content Heuristics
