An Axiomatic Framework for Result Diversification

Informally speaking, diversification of search results refers to a trade-off between relevance and diversity in the set of results. In this article, we present a unifying framework for search result diversification using the axiomatic approach. The characterization provided by the axiomatic framework can help design and compare diversification systems, and we illustrate this using several examples. We also show that several diversification objectives can be reduced to the combinatorial optimization problem of facility dispersion. This reduction results in algorithms with provable guarantees for a number of well-known diversification objectives.