The Solution

     The problem of knowledge representation is fairly complex and the Roche-knDB solution is based on the use of logic and inference methods in conjunction with knowledge representation structures, with the final goal being the capability to infer new knowledge from existing knowledge.

     The Kernel of the system is the Lisp/PROLOG logical reasoning model.

     The representation of the relation between the Information elements in the system and their instances is based on the two top of the art approaches widely used by various Artificial Intelligence (AI) models: Predicate Calculus and Bayesian (Belief) Networks. The Resource Description Framework (RDF) is a foundation for processing metadata and it is used in conjunction with the inference methods of Predicate calculus to build a tree like structure of Information elements, which represents the backbone of the knowledge base system.

     With the introduction of Bayesian Networks (BN) a new dimension is added to the Tree like Taxonomical structure, which will allow the users to find relations between Information clusters situated in completely independent branches of the Tree like structure. Furthermore the user is able to make valid conclusions about the relation of concepts (based on instances of information elements) to certain documents and thus locate existing documents referring best to a certain problem, subject, etc. The resources of the knowledge base is stored in an object oriented relational database and the instrument (interface) to extract/store information from/in the database is a Lisp module (fQL) which is based on the principles of “Predicate Language" and Bayesian Networks.