The article is devoted to the issues of linguistic and mathematical support for automatic identification of cognitive-communicative actions that are realized in two speech genres of a scientific empirical text - new phenomenon description and experimental data classification. The authors consider that cognitive processes concealed from direct observation are objectified in genre forms, and the latter are such link in the chain of cognitive-communicative activity that allow to systematize and formalize these processes. The category of speech system, a basic one in functional stylistics and some areas of discourse analysis, is explicated as a genre-related level of abstraction. A speech act is considered not activity-isolated but as its intrinsic element. We pro-pose a method for identifying mental actions objectified in a text that supposes the procedures of automatic linguistic text analysis, matching heterogeneous semantic networks of sentences with contextually free patterns, and also specifying the con-tent of the performed actions using sequence classification. The results of automatic identification of cognitive-communicative actions realized in the texts of these genres are presented.