Consolidating diverse user profiles based on the profile models of adaptive systems

Abstract

Profile-based adaptivity is an important ingredient of interactive systems. Today, although users keep many profiles in different applications, adaptive systems still request them explicitly. While lingua franca methods on profiles are suggested, unless standardized, they are hardly deployed by different vendors. We present an approach to consolidate diverse user profiles based on a profile model that is supplied as input. The latter is instantiated in our Gandalf system, where user profiles from various sources are aggregated, merged and mapped to any given model, by also preserving private user attributes. No common models for profiles are assumed, neither any shared models across adaptive systems are prescribed. Our method uses a thesaurus service, while it proposes lightweight rules for structure matching and conflict resolution to accompany the input profile model. Gandalf is under implementation as a web service, and allows adaptive systems to hook custom pre- and post- processing logic on profiles using JavaScript.