Agent-Based Generation of Personal Thesaurus

Grzegorz Popek, Radoslaw P. Katarzyniak

Abstract: An agent autonomously develops thesaurus which is an opposite to a typical situation in which agents are delivered with a thesaurus proposed by a designer. The agent uses concepts which emerged from the process of semiosis. These concepts are modeled as values of linguistic variable, and therefore, a single fuzzy variable can be correlated with each of them. Also, they can be paired with linguistic hedges (very, slightly, etc.). An agent finds types of relations between concepts (very, slightly, less general than, ...) based on presented commonsense requirements and a similarity between different concepts and combinations hedge- concept. In the end, all relations are gathered together by a proposed algorithm to compose a private self- developed thesaurus.