Defining the dynamicity and diversity of text collections

Abstract

In Information Retrieval collections are often considered to be relatively dynamic or diverse, but no general definition has been given for these notions and no actual measure has been proposed to quantify them. We give intuitive definitions of the dynamicity and diversity properties of text collections and present measures for calculating them based on the notion of novelty. Experimental results show that the proposed measures are consistent with the definitions and can distinguish collections effectively according to their dynamicity and diversity properties. © 2010 Springer-Verlag Berlin Heidelberg.