Uncertainty in Web Data

Knowledge representation in very distributed environments, such as the Web, arise reliability and mutual consistency concerns. Starting from an exposition of current Web knowledge representation languages, this work propose solutions for various kinds of uncertainty in web data. A first kind of uncertainty arise from inherently vague knowledge. For this reason RDF and RDF Schema are extended with a fuzzy syntax and semantics, in the same way as fuzzy logic allows the representation of uncertain knowledge not expressible in first-order logic. The extension is then further generalized to arbitrary ¿annotations¿ on web data, where annotations are taken from any partially-ordered set. Another kind of uncertainty arise from contradicting information. Contradictions are harmful for traditional reasoners, as entailment is not usable in presence of inconsistency. Two of the possible solutions analyzed and extended are to restore consistency in the original knowledge base, and to use non-standard reasoning allowing to derive meaningful answers even with inconsistent inputs.