I've been playing around a little with MySidekick, a new (in beta) search engine which uses queries from previous searches (from all users of the system) to assist in new searches:
Our new social search engine is called mySidekick (www.mySidekick.com) and it harnesses the inherent efforts of potentially millions of daily Internet searchers to provide better results for everyone. Here's how it works: When most people search for something, they end up changing their search request about 3-4 times before finally finding what they are looking for in the results. Using Yahoo to provide our underlying search results, our new system automatically tags the sites you find with the terms you entered across related queries in your search session. In this way, the system learns the best tags for each site on the web as these sites are found and automatically tagged by other people. The end result is a set of more relevant search results that we list as the "People's Choice" results on top of the regular search results. The system is easy to use and you don't even have to do anything different than what you already do to search at Google or Yahoo.
At a high level, I'm interested in this approach. However, the system suffers from something of a cold-start problem. Being the first to enter each search (which, for early users, is likely to be the case) means that one is getting results very similar - or identical - to those provided by the search engine backing the system.
Comments