Clustering XML documents using structural summaries

Abstract

This work presents a methodology for grouping structurally similar XML documents using clustering algorithms. Modeling XML documents with tree-like structures, we face the 'clustering XML documents by structure' problem as a 'tree clustering' problem, exploiting distances that estimate the similarity between those trees in terms of the hierarchical relationships of their nodes. We suggest the usage of tree structural summaries to improve the performance of the distance calculation and at the same time to maintain or even improve its quality. Experimental results are provided using a prototype testbed. © Springer-Verlag 2004.