Keywords: neural network text analysis, document anomaly index, sigmoid normalization, weighting coefficients, loss function
A Method for the Automated Detection of Anomalies in Educational Documents Based on Neural Network Text Analysis
UDC 004.89:371.26
This article examines a method for the automated detection of anomalies in educational documents based on neural network text analysis. The proposed approach employs semantic, structural, and stylistic analysis to evaluate the quality of academic assignments. To this end, an integral anomaly index is introduced, enabling a quantitative assessment of the extent to which a document deviates from standard requirements. The principles of calculating this index are discussed, including weighting coefficients and result normalization methods. It is demonstrated that the use of sigmoid normalization enhances the interpretability of the final scores. The proposed method can be applied in systems for the automated monitoring of learning outcomes and for providing intelligent support to instructors during the assessment of academic work.
1. Шамарина Е.А. Сентимент-анализ на основе нейросетей-трансформеров / Е.А. Шамарина, А.И. Гусева, В.С. Киреев // XXV Международная научно-техническая конференция «Нейроинформатика-2023»: Сборник научных трудов, Москва, 23–27 октября 2023 года. – М.: Национальный исследовательский ядерный университет «МИФИ», 2023. – С. 292–301.
2. Усен Е.Б. Применение нейросетевых моделей для семантического анализа публикаций в социальных сетях / Е.Б. Усен // Вестник науки. – 2025. – Т. 4, № 2 (83). – С. 480–485.
3. Иванов В.М. Интеллектуальные системы: учебное пособие для вузов / В.М. Иванов; под научной редакцией А.Н. Сесекина. – М.: Издательство Юрайт, 2025. – 88 с.
4. Кудрявцев В.Б. Интеллектуальные системы: учебник и практикум для вузов / В.Б. Кудрявцев, Э.Э. Гасанов, А.С. Подколзин. – 2-е изд., испр. и доп. – М.: Издательство Юрайт, 2025. – 165 с.
Keywords: neural network text analysis, document anomaly index, sigmoid normalization, weighting coefficients, loss function
For citation: Nazarov A.A. , Preobrazhenskiy A.P. , A Method for the Automated Detection of Anomalies in Educational Documents Based on Neural Network Text Analysis. Bulletin of the Voronezh Institute of High Technologies. 2026;20(3). Available from: https://vestnikvivt.ru/ru/journal/pdf?id=1504 (In Russ).
Received 02.07.2026
Revised 07.08.2026
Accepted 07.08.2026