Preference Functional Dependencies for Managing Choices

Abstract

The notion of user preference in database modeling has recently received much attention in advanced applications, such as personalization of e-services, since it captures the human wishes on querying and managing data. The paradigm of preference-driven choices in the real world requires new semantic constraints in modelling. In this paper, we assume preference constraints can be defined over data domains and thus the assumption gives rise to preference relations as a special case of ordered relations over schemas consisting of the preference, preferencedependent and preference-independent attributes. We demonstrate that Lexicographically Ordered Functional Dependencies (LOFDs) can be employed to maintain the consistency of preference semantics embedded in preference database, since prioritized multiple preferences can be represented. We thus define a useful semantic constraint in terms of a set of LOFDs, called Preference Functional Dependencies (PFDs), in order to capture the semantics of the preference ranked data. We exhibit a sound and complete axiom system for PFDs, whose implication problem is shown to be decidable in polynomial-time. We also confirm the existence of Armstrong preference relations for PFDs, a fundamental result related to the practical use of PFDs in database design.

This work is partially supported by RGC CERG under grant number HKUST6185/03E.