This paper describes the concepts of architecture of modern systems for e-learning based on semantic technologies with the aspects of adaptation, personalization and dynamic adaptability of learning material. The proposed architecture is able to generate the student's profile based on basic and historical data on the student which ensures the personalization of the system. The personalized student data enable us to ensure dynamic generation of the course in the system according to individual needs of the student. The effects of adaptation are achieved by adapting the system according to the categories of adaptation. The paper also gives an overview and description of the operation of each individual component of the system.