Progressive ranking based on a dominance list

Preference queries aim at increasing personalized pertinence of a selection. The most famous ones are the skyline queries based on the concept of dominance introduced by Pareto. Many other dominances have been proposed. In particular, many weaker forms of dominance aim at reducing the size of the answer of the skyline query. In most cases, applying just one dominance is not satisfying as it is hard to conciliate high pertinence, i.e. a strong dominance, and reasonable size of the selection. We propose to allow the user to decide what dominances are reliable, and what priorities between those dominances should be respected. This can be done by defining a sequence, eventually transfinite, of dominances. According to that sequence, we propose operators that compute progressively the ranking of a dataset by successive applications of the dominances without introducing inconsistencies. The principle of progressive refinement provides a great flexibility to the user that can not only dynamically decide to stop the process whenever the results satisfies his/her wishes, but can also navigates in the different levels of ranking and be aware of the level of reliability of each successive refinement.