Preference Representation And Revision In Intelligent Information Processing

The binary relation framework has been proved to be applicable tomany real-life preference handling scenarios. In this paper, the authordevelop methods of representing preferences in the binary relationpreference framework. The preferences refer to the binary relations in strictpartial order(SPO). The preference framework allows for finite and infinitepreference relations. In order to make preference representation andrevision easier, the following research directions pursued.Firstly, we make a study of the preservation of SPO in preferencerelations. Preference contraction refers to the operation of revising apreference relation by discarding its subset. All the contraction operatorsdeveloped here satisfy a common property of operators of preferencerevision– they preserve SPO axioms of a modified preference relation.Thus, preference relations are closed in the operation of preferencecontraction. The fundamental principle of preference contraction is thecontraction should be minimal. In order to contract a preference relation bya base contraction, it discards a minimal full contractor of base contractor,whose removal from the preference relation is needed to satisfy the SPOpreservation property.Secondly, we propose an operator called preference-protectingminimal contractor to give a user more freedom do specify preferencerevision more precisely. It is a constrained version of the minimalcontraction operator. The operator has an addition parameter– a subsetP+ of the original preference relation which has to be protected. Such P+may be viewed as the set of preferences which are of the highestimportance for a user.Thirdly, we propose meet contraction. In some cases, in order tospecify the preferences which need to be protected after contraction orelaborate the contraction, a user may want to know which preferences maypotentially be lost when discarding a base contraction.Finally, we also study some preference query optimization techniqueswhich can be used in the presence of contraction. We evaluate the proposedalgorithms experimentally and present the results.