Empirical Analysis of the Rank Distribution of Relevant Documents in Web Search

This paper proposes an empirical approach for analyzing the rank distribution of relevant documents in Web search. From a methodological point of view a new transaction log analysis method is proposed; the relevance of documents is studied over transaction sessions rather than single transactions. From a practical point of view the paper provides insights about the actual rank distribution of relevant documents in Web search, with several consequences for the design of applications related to Web search.