Hasso-Plattner-Institut20 Jahre HPI
Hasso-Plattner-Institut20 Jahre HPI
  • de

Dissertation Rehab Alnemr

Computer scientists strived for years to capture features of our social communities to reach a point where mutual understanding and knowledge sharing are possible. A critical social concept that contributes greatly in decision making is reputation. In open environments, reputation management approaches are used to establish trust relationships between unrelated parties, especially if enforcement methods like institutional policies are not implemented. They provide means for collecting reputation data about entities, computing reputation values, and performing actions based on these values to determine trustworthiness and make automated decisions. Each reputation system has its own method to query, store, aggregate, infer, interpret and represent reputation information, hence locking reputation interactions within isolated silos. In most reputation systems, the reputation context is not embedded within the given reputation representation. Every domain has its own information sources as well as its own requirements. Therefore, the representation -not the calculation- of reputation should be unified between communities in order to facilitate knowledge exchange.

In this thesis, we focus on the analysis of reputation representation and its expressiveness in the context of online communities and service oriented environments. We conduct user studies to analyze several aspects of reputation. The results showed that current representations of reputation are not sufficient to convey an entity's reputation and that users always use more information than the displayed ratings in taking their decisions. These are part of the reasons of why current reputation is not portable i.e. can not be exchanged between different communities. To acquire a reputation that is interoperable in nature, generic reputation ontologies have to be developed. In this dissertation we present a generic and interoperable reputation representation model, which we call reputation object, developed using semantic technologies. In the first part of the thesis, we present a review of the state of the art research on trust management and reputation systems followed by a background on semantic web technologies. In the second part, we explore the challenges addressed in this thesis and describe our user studies. Next, we demonstrate our reputation object ontology model and describe how the use of semantic technologies leverage our goals. We then extend our model by presenting a context-aware reputation framework which enables reputation portability and defines its main component services. Next, we describe our implementation and how it can be integrated in other systems. Since the value of reputation can only be seen within a system or a working domain, we evaluate our work through several use cases and discuss other possible use cases.