Contribuţii la problematica achiziţiei de cunoştinţe pentru agenţi inteligenţi

Abstract

The human society is a complex system. The knowledge is a creation of the society, and in turn, it creates the society. A man's knowledge development occurs through social interaction, since their infancy. Social interactions have multiple dimensions and argumentation is an essential piece from the puzzle. Through argumentation, people reinforce and convey knowledge. Argumentation is an essential tool for knowledge acquisition, regardless of the type of entities involved. The thesis main idea consists in applying specific argumentation mechanisms to intelligent agents in order to acquire new knowledge. Classically, the knowledge acquisition problem comes from the knowledge formalization and implementation in a computer system. Historically, there have been specified and implemented many systems and formalisms that enable the acquisition of knowledge from an expert. The acquisition of knowledge is the first step in the development of an expert system. In multi-agent systems, the agency already has prior knowledge extracted with classical methods of knowledge discovery on large data sets. The thesis identifies the difficulties of extending the knowledge base through classical methods and propose a novel approach based on argumentation, approach that allows self-evolvement of agents without human expert intervention.