Knowledge agents via logic programming and fuzzy reasoning

Implementing knowledgeable applications benefits from the ability to model problems on the knowledge level. In practice, this usually means engagement with the logical description of problems. However, when (crisp) logic is applied as a tool in application design, modelling the fundamental vagueness of the human knowledge may soon become an obstacle. We claim that the two worlds of logic programming and fuzzy reasoning should coincide when practical knowledge applications via logic programming are concerned. In this thesis, we point out and analyse an approach that provides the necessary methodological and technical means achieving this goal. In particular, we consider the various aspects of systems called fuzzy knowledge agents. By a fuzzy knowledge agent we mean a tool that helps users to process and manage information via the logical description of the domain, benefiting from the use of fuzzy models when applicable. The chief utility of fuzzy knowledge agents lies within their ability to formulate and encapsulate information through logical and linguistic means, providing a logic-based end-user interface to the underlying information on the knowledge level. In this thesis, we establish and analyse a reasoning architecture that allows realising the main steps of designing and implementing fuzzy knowledge agents, including: inception and elaboration of the domain vocabulary and the associated logical procedures (to be captured with type-1 fuzzy logic programs); appropriate modelling of the domain concepts, based on heuristic and statistical arguments (to be modelled with fuzzy sets, induced from empirical data when appropriate); and construction of the reasoning and query applications. In addition, we consider the concept of context-aware logic programs and review the necessary technical components of fuzzy knowledge agents. Finally, we evaluate the methods and discuss their applicability with several illustrative use cases.