Single-pass clustering for peer-to-peer information retrieval
Iraklis Klampanos, Joemon M. Jose, C. J. van Rijsbergen · 2006
Document clustering has been a particularly active research field within the Information Retrieval (IR) community. Among the numerous clustering algorithms proposed, single-pass clustering stands out in terms of both time and space efficiency. However, it is generally acknowledged that single-pass clustering has a major defect, namely its output depends on the order in which documents are presented. Building on our previous work, and having identified single-pass clustering as potentially useful for P2P IR, we study the extent to which this is true in practical terms. We do so by experimenting with two large web-based testbeds, which are suitable for Peer-to-Peer IR evaluation. The results of our study show that document ordering does not practically matter for single-pass clustering.