On the Diversity Within Internet Search Engines : A Clustering Approach Using

We present experimental results on how internet search engines compare with each other based on their query results. If the relationship between two rankers is substantially different for two different queries, it suggests that the queries are of a different nature. We use this observation to also investigate how queries relate to each other based on the relative performance of different internet search engines on them. Specifically, we show how clustering approaches can reveal relationships among search engines, and relationships among queries modulo those search engines. Our results have implications on aggregation of search results and on reverse engineering of ranking formulas used by search engines.