A Divide-and-Merge Methodology for Clustering

Abstract

We present a divide-and-merge methodology for clustering a set of objects that combines a top-down ``divide'' phase with a bottom-up ``merge'' phase. We use an efficient spectral algorithm for the divide phase and a simple strategy for the merge phase. We present a meta-search engine that uses this technology to cluster results from web searches. We also give emperical results on text-based data where the algorithm performs better than or competitively with well-used existing methods.

Back to publications
Back to previous page