A new way of searching the web!
I was recently looking for information on the risk ratings of the various GGs and what initiatives exist to study these phenomena. It took me over half an hour to find what I was looking for. The problem was simple - too much information. Try googling on Tsunami and you get well over 11 mill results. How does one go over this result set and find relevant information?
Enter Search Result Clustering - a non-linear way of looking at results as compared to the traditional "ranked list" that makes it faster for the user to reach his desired results. The idea is to use a clustering algorithm to aggregate related results together under derived headings. Want to see how this works? Check out http://wsm.directtaps.net/default.htm - courtsey MSR Asia. Download the toolbar or search from the site directly!
Note that this is not a search or indexing technology - the input to the algorithm is the original query and the returned ranked list of results (could be from any search engine) - the algorithm extracts phrases from results-titles and ranks these phrases as cluster headings. More details on http://research.microsoft.com/users/hjzeng/p230-zeng.pdf.
