Abstract :
There are at least three possible ways that documents
are distributed by relevance: informetric (power law),
inverse logistic, and dichotomous. The nature of the
type of distribution has implications for the construction
of relevance ranking algorithms for search engines, for
automated (blind) relevance feedback, for user behavior
when using Web search engines, for combining of outputs
of search engines for metasearch, for topic detection
and tracking, and for the methodology of evaluation
of information retrieval systems.
