Probabilistic ontology for social background verification: A belief fusion use case

The information explosion and subsequent problem of information overload has become more prominent in recent years. The situation can be aptly described by John Naisbitt's quote, “We are drowning in information but starved for knowledge”. The information overload in intelligent systems has highlighted the restricted ability of knowledge representation (KR) formalisms to cope with various forms of uncertainty in knowledge. Probabilistic extensions of KR formalisms are the most investigated for uncertainty representation. Probabilistic extensions of formal logics offer expressiveness at the cost of tractability of reasoning. On the other hand Probabilistic extensions of graphical models makes simplifying assumptions to achieve better computational performance but are less expressive. In fact one of the most noted problem of probabilistic KR formalisms is the difficulty to achieve the expressiveness-of-representation vs tractability-of-reasoning trade off. In addition the KR formalism for open domains such as Semantic Web need to facilitate information sharing and semantic interoperability as well. Ontologies have already been successfully used as KR formalism as they facilitates semantic interoperability and therefore automated reasoning; one of the essentials to cope with information overload. However traditional ontologies cannot deal with uncertain knowledge. Probabilistic ontologies use probabilistic KR formalisms and therefore can cope up with domain uncertainty. This paper investigates the applicability of probabilistic ontology as a KR formalism for a Semantic Web use case. Previous work on probabilistic ontologies has investigated use cases such as fraud detection, threat detection, and breast cancer detection in government, military and healthcare domains but not in Semantic Web. Uncertainty also manifests in various forms and all of them are often not present in single use case. We have chosen the use case for achieving belief fusion under uncertainty to investigate how probabilistic ontology copes with all three expectations for a semantic web KR formalism; expressiveness, tractability and semantic interoperability. Proposed probabilistic ontology based system automates the background verification part of human recruiting process. This use case is a representative case for belief fusion and asserts Bayesian networks as belief fusion operator.