The methodology has been researched and methods, models, algorithms have been developed for implementing the technology for increasing the reliability of information in electronic document management systems with mechanisms for determining the information proximity of elements, code words, by bit, digital, linear, modular summation, using a multilevel morphological and n- gram structured model for analyzing texts on natural language. Computational schemes have been developed for describing sets, basic functions for initializing all objects, using logical and statistical relationships of text elements, forming a knowledge base with sets of rules, and information control statements. A technique has been developed, which is implemented in the form of a Python program for estimating the parameters of mechanisms for increasing the reliability of information. Mechanisms based on binary modular summation of information bits of a code word over lines of text are proposed. The value of the least non-negative residue of the weight of the information vector, estimates of the probabilities of undetected errors, coefficients of gain in the reliability of information, labor intensity and cost of information control are obtained. A software package for improving the reliability of information in the C++ language has been implemented, in which the proposed mechanisms are synthesized in the environment of the CUDA parallel computing technology. The software package detects and corrects multiple errors in textual information.