Currently, due to the successful use of neural network technologies for analyzing data of various formats, the range of problems that can be solved using mathematical modeling methods has significantly expanded. The paper deals with the topical task of analyzing speech perception by social media data and assessing the level of conflictogenity, social approval / stress of the residents regarding urban planning projects. The purpose of the paper is to develop and show on a practical example the effectiveness of the integration of neural network and mathematical models for solving such tasks. Mathematical models are built using the methods of mathematical statistics and topological data analysis. © 2021, International Institute for General Systems Studies. All rights reserved.