Sparse Spatial Selection for Novelty-based Search Result Diversification

GIL COSTA G. VERONICA; RODRYGO SANTOS; CRAIG MCDONALD; IADH OUNIS

Conferencia; String Processing and Information Retrieval.; 2011

Novelty-based diversification approaches aim to produce adiverse ranking by directly comparing the retrieved documents. However,since such approaches are typically greedy, they require O(n2) documentdocumentcomparisons in order to diversify a ranking of n documents. Inthis work, we propose to model novelty-based diversification as a similaritysearch in a sparse metric space. In particular, we exploit the triangleinequality property of metric spaces in order to drastically reduce thenumber of required document-document comparisons. Thorough experimentsusing three TREC test collections show that our approach isat least as effective as existing novelty-based diversification approaches,while improving their efficiency by an order of magnitude.