Summary: Benefits of Semantic Web technologies for knowledge modeling and reasoning are well established. However, there are still some serious deficiencies to deal with uncertainty, which is an essential requirement for many nowadays applications. This article presents a framework for semantic beliefs fusion. It provides means for the representation of uncertain ontological instances and offers a way to reason on this knowledge. Uncertain instances can have both uncertain classes and properties. Different sources populate the same ontology, according to their own state of belief. The more reports of the same uncertain phenomenon we will collect, the more likely a precise and accurate description of this phenomenon will be obtained. The Evidential theory is used to fuse that uncertain semantic information. For that, notions of semantic inclusion and disjointness between ontological instances are introduced.