The paper presents an approach for the dynamic generation of a complex structure of personalized web pages for learning purposes, reflecting the ontology of the considered domain. The need of assuring a holistic character for the body of knowledge induced in the learner's mind is emphasized. This is very important in the learning processes, especially nowadays, in the context of the huge amount of information available on the web and of its permanent evolution. The approach permits the adaptation of the content of the generated web pages to the incoming information from the web. New information is extracted, annotated and coherently integrated in the body of knowledge in order to keep the holistic character of the body of knowledge. Personalization is achieved by filtering the semantic network according to the learner model, which keeps the list of concepts known or unknown by the learner. The approach was used in an EU INCO Copernicus project for computer aided language learning.