PISA: A framework for integrating uncooperative peers into P2P-based federated search

摘要

In the past years, federated search over peer-to-peer (P2P) networks has attracted considerable attention from the information retrieval community. Most previous work assumes a so-called cooperative environment, where each peer can actively participate in information publishing and document indexing in a distributed manner. In contrast, little prior work has addressed the problem of incorporating uncooperative peers, which do not publish their own corpus statistics over a network. In this paper, we present a P2P-based federated search framework named PISA, which incorporates uncooperative peers when providing search service. Specifically, we (i) propose a heuristic query-based sampling approach named HQ.BS, which can obtain high-quality resource descriptions from uncooperative peers at a low communication cost; (ii) present two result merging methods, called RISE and R1SE+, to merge the results returned by uncooperative peers; (iii) develop a method called Controlled and Selective Update (CSU) to efficiently maintain the index directory of PISA. Our extensive experiments on the TREC WTlOg data set demonstrate that PISA can provide high-quality search results and integrate uncooperative peers at low cost.