In this contribution we are going to explore both syntactic and semantic structures of the dictionary definitions as they can be found in standard printed or machine dictionaries (as e.g. SSJC or NODE). Particularly, we would like to pay attention to the usual ways of building dictionary definitions, and to examine whether and to what extent they reliably and systematically follow the standard principles of the formal description of language data. The results following from the presented analysis should be applied to building dictionary definitions (glosses) for Czech WordNet (being further developed in the frame of Balkanet Project) and also the new Czech Lexical Database.