Evaluation of text clustering algorithms with n-gram-based document fingerprints

Abstract

This paper presents a new approach designed to reduce the computational load of the existing clustering algorithms by trimming down the documents size using fingerprinting methods. Thorough evaluation was performed over three different collections and considering four different metrics. The presented approach to document clustering achieved good values of effectiveness with considerable save in memory space and computation time. © Springer-Verlag Berlin Heidelberg 2009.