The paper presents an adaptation of the model of deep text analysis elaborated earlier by Yu. V. Popov and T. P. Tregubovich for German. The development of the model for Chinese is based on principles of the general linguistic predicational conception of V.A. Kurdyumov, which nowadays is the only one taking into account the specificity of isolating languages. The advantage of the model, presented in the article, consists in revealing the mechanism of Chinese text production and presenting the functioning of cohesion phenomenon not only in the micro-, but also in the macrostructure of text. The algorithm of analysis, proposed in the article, has a certain degree of universality, and, with some modifications, it can be implemented in text analysis of different types of languages. It is concluded that with additional elucidations for several steps of the algorithm it is possible to use the proposed model of deep text analysis both for monologic and dialogic texts.