David Plotnikoff writes about the need for a teachable search engine.
My teachable search engine would watch my every move, looking at sets of search results and collecting a wealth of data about the pages I view — and perhaps those I don’t, as well. It would archive this vast personal history but it would not be able to tie it to my personal identifying information. In a perfect world it would only know me when I log in as “anonymous user No. 12345.”
The search engine would note whether I visited a page once — or 20 times. It would find similarities between relevant pages, without me ever having to tell it, “This page was close to what I’m looking for. This one was not.” And in this manner, over the span of a hundred or a thousand searches, it would build a pretty good understanding of how I search. With each new search it would turn to this history and look for similarities between the current query and all previous searches.
A good place to try out some of these ideas is with a Blog Search Engine. Its a much narrow cluster of documents, and its also about “connecting” to people like us.