Where and How Duplicates Occur in the Web

Abstract

In this paper we study duplicates on the Web, using collections containing documents of all sites under the .cl domain that represent accurate and representative subsets of the Web. We identify duplicate and near-duplicate documents in our collections, studying the distribution of documents in clusters of duplicates. We also study the occurrence of duplicates in both parts of our Web graphs - connected and disconnected component - aiming to identify where duplicates occur more frequently. We originally show that the number of duplicates in the Web is expressively greater than the number of duplicates in the connected component of the Web graph. Works that previously estimated the number of duplicates in the Web used collections of connected components of the Web. In those cases the sample of the Web was biased