Reliability of answers in semantic peer-to-peer networks

Overview & Synopsis

This thesis improves the reliability of answers in semantic peer-to-peer networks in different ways. The first part of the thesis focuses on the problem of returning only well-founded answers of a query when the global network of such a system is inconsistent. Two algorithms have been proposed. The first one detects causes of inconsistencies. It terminates, is decentralized, complete and correct. The second algorithm uses the results of the first one in order to make only well-founded reasoning. It terminates, is decentralized and correct. Promising results have been obtained from experimentations. The second part of the thesis considers the problem of modeling and handling peers' trust into returned answers. Based on a probabilistic setting, the proposed model of trust has a clear semantics and trust towards answers can be estimated by peers with a good precision using few observations.