The tantalizing new prospect of index-based diversified retrieval
George Tsatsanifos · 2013
In this paper, we propose efficient algorithms for result diversification over indexed multidimensional data. We develop algorithms under the prism of a centralized approach, as in a database. Specifically, we rely on widely used multidimensional indexes, like the R-tree. In principle, our schemes adopt a maximal marginal relevance (MMR) ranking strategy and leverage interchange and greedy diversification techniques. Hitherto, mostly combinatorial aspects of this problem have been considered which require scanning the entire data, and therefore, existing solutions are costly.