Development of an automated text data preprocessing module with adaptive control for neural network models

Abstract

The article discusses the development of an automated data processing system module for preparing datasets before training neural networks. The aim of the work is to overcome the problems associated with low quality and heterogeneity of input data, which lead to reduced productivity and retraining of models. The article proposes a comprehensive module architecture that includes a formal decomposition of the preprocessing process, a multi–level control loop for operational regulation and adaptation, as well as a detailed data model and database schema based on the MariaDB database management system. Special attention is paid to the methods of text information processing in NLP tasks. The presented solution makes it possible to standardize and significantly speed up the data preparation stage, minimizing manual labor and the influence of the human factor. The results of the work will be useful to researchers and developers seeking to improve the quality, reproducibility and efficiency of machine learning models through reliable automation of data preprocessing. doi 10.54708/19926502_2026_30211241