An Adaptive Sorting Technique for the Web
Web-sites such as news.google.com can provide a more personalized
presentation to visitors by applying a simple sorting technique.
Long lists of information--principally of Web-links--can easily be
sorted by probable interest so that links a user may be most
interested in gravitate toward the top of the page.
Google's news page is a good example to explain how to apply this
Google's news page lists news article links from many disparate
yet related sources--from traditional new sources such as Time
and CNN to online news such as Salon.com and Slashdot.org. The
list is sorted in (what appears to be) an arbitrary manner.
There are two levels to Google's sorting approach; links are
categorized first--Political, Entertainment, Science, Sports, etc.
Each category is sorted by publication time, latest first. What
Google has (intentionally or not) basically applied a weighting
factor to each link: Category and Time. By further weighting with
keywords--something Google has capability of already--Google can
simply maintain a selection history for each unique user, and use
it to sort by this weighting factor.
For example, if we were to look at my viewing history, one would
find that I rarely view articles categorized as Sports and
Entertainment, and mostly view articles categorized as Science
and Politics. If there were a list of keywords attached to each
article-link Google would have a measure of what kind of articles
I mostly view.
Google can then sort the articles list to my probable liking--articles
most likely of my interest at the top.