From 'Identical' to 'Similar': Fusing Retrieved Lists Based on Inter-document Similarities

摘要

We present a novel approach to fusing document lists that are retrieved in response to a query. Our approach is based on utilizing information induced from inter-document similarities. Specifically, the key insight guiding the derivation of our methods is that similar documents from different lists can provide relevance-status support to each other. We use a graph-based method to model relevance-status propagation between documents. The propagation is governed by inter-document-similarities and by retrieval scores of documents in the lists. Empirical evaluation shows the effectiveness of our methods in fusing TREC runs.