This post originated from an RSS feed registered with Java Buzz
by Russell Beattie.
Original Post: Personalized Search Thought
Feed Title: Russell Beattie Notebook
Feed URL: http://www.russellbeattie.com/notebook/rss.jsp?q=java,code,mobile
Feed Description: My online notebook with thoughts, comments, links and more.
I stop my mad dash to finish up all this stuff for DEMO to write about an idea I've had and haven't mentioned concerning Personalized Searches. I'm not a search or aggregator guru like the Feedsters and Technorati's of the world, but it seems to me a really obvious use of the Google API is personalized searches.
There seems to be some scratching at the surface of the idea, but nothing I've seen really learns what you like just yet. To me it seems pretty simple - put an interface in front of Google where you have to have an account to use it. The account requires some personal details like where you live, etc. Then as you search, the personalization service "wraps" the resulting links so that as a user clicks on those links, the service can learn from which ones were accurate (sending back a redirect to the original link back to the user). The search service also could fire off a HTTP get of its own to add that page to a Bayesian database of some sort... per search "session" the service would record all the pages that the user clicked on, and disregard all except the very last page. Because the last page you click on is probably the one where you found the information you wanted, right?
Now, as people continue to use the service the service can start to learn what to add to the query params before passing it off to Google, and maybe filtering the return results. Looking for "java"? Well, if you're a programmer, you get back pages having to do with development, if you're a frequent traveller, you get travel stuff to the island, etc.
Now, maybe someone is doing this and I don't know about it, but it seems like a pretty doable sort of thing. Maybe even as an offline tool - this could be what Microsoft is planning for their new MSN search, which supposedly has client integration. Again, the idea is to make the personalized searches as scalable as possible, so all your little Bayesian filter is doing is adding additional parameters behind the scenes like "+San +Francisco" or adding an additional layer of filtering before the results get to you. This last part is probably a lot like the "Safe Search" child-filters do...
Okay, done with the brain dump. Back to work... (Someone leave me a comment if this is out there and I don't know about it. I think it'd be interesting to try out.)
Previous Topic
