Integration of modal and fuzzy methods for agent's knowledge representation

An application of the natural language in articial systems is not a new idea and follows from a natural need for an improvement of a process of communication between a man and a machine. A system { the machine { can be either a receiver or a source of messages. The thesis focuses on the latter and deals with a task of a generation of semantic messages in an arti cial system. This research follows a pragmatic example for an application of the natural language in articial systems. The example consists of a system which provides textual summaries of numerical data. In particular, an application scenario focusing on a provision of textual description of city-tra c states to an external user of a system is investigated. The research focuses on situations with incomplete observational data in which case the information about said incompleteness is aggregated and conveyed to the user. In the past a task of describing the data with words and a task of introducing the notion of sign of a language into articial systems have been considered separately. Proposed solutions follow two methodologies known in an area of Articial Intelligence. The first methodology is related to the modelling of a binding between numerical data and concepts of the natural language. It proposes to use a fuzzy-linguistic (FL) model introduced by Zadeh in 1975. The second methodology deals with the problem of grounding i.e. the problem of introducing notions of sign and meaning into an articial system stated by Harnad in 1990. In particular, it follows a branch of the grounding-research which focuses on grounding of modal statements with autoepistemic operators of possibility, belief, and knowledge. The theory of grounding enables articial agents to deal with incomplete knowledge and ground autoepistemic operators in their experiences. The main accomplishment of an undertaken research is an integration of a fuzzy-linguistic approach to modelling concepts with a theory of grounding of modal statements with autoepistemic operators of possibility, belief and knowledge. The author integrates the grounding of a fuzzy-linguistic variable with existing mechanisms of grounding of modal statements in articial agents. In particular, the thesis shows that it is possible to design an articial agent equipped with a hybrid mechanism of grounding, which fulfills commonsense requirements known from the natural language discourse.