Various approaches to solving abstracting automation problems are analysed. Formalisation of the process of building the structure of scientific and technical texts is considered. A text structure correctness criterion has been proposed and the corresponding structural characteristics and limitations have been introduced.