In addition to the top first document which should be quite relevant to the query, the following top documents should also satisfy two constraints:
1. They should be as similar as possible to the query.
2. They should be as dissimilar as possible from each other.
It can make IR "result diversity", that is also the reason we need R(d|Q,U,{d1…di-1}), where d1..di-1 are documents already shown to the user.
--Shu Wang
