TY - CONF

T1 - Automatic Recognition of Historical Documents of the Beginning of the 20th Century in Russian

AU - Barakhnin, Vladimir

AU - Gudkov, Stepan

N1 - V. Barakhnin and S. Gudkov, "Automatic Recognition of Historical Documents of the Beginning of the 20th Century in Russian," 2025 IEEE XVII International Scientific and Technical Conference on Actual Problems of Electronic Instrument Engineering (APEIE), Novosibirsk, Russian Federation, 2025, pp. 1-4, doi: 10.1109/APEIE66761.2025.11289354. Research carried out under the state contract with ICM&MG SB RAS FWNM-2025-0005.

PY - 2025/11/14

Y1 - 2025/11/14

N2 - The large number of undeciphered handwritten historical documents stored in archives forces historians to precede the study and introduction of these documents into scientific circulation by manually deciphering them – a very labor-intensive process. The use of information technology makes it possible to automate this process. The work is aimed at creating a system for automatic recognition of books of decisions of volost courts based on a hybrid approach: neural network methods for recognizing handwritten text and linguistic analysis methods for correcting recognition errors. The paper examines the development of one of the system modules: a letter recognizer. An ensemble approach based on convolutional neural networks is used, which allows achieving an F1-score of 75.1%. The first paragraph – the introduction – sets out the problem, the second provides an analysis of the applicability of existing recognition systems to this problem, and the third describes the development of a specialized system for recognizing books of decisions of volost courts.

AB - The large number of undeciphered handwritten historical documents stored in archives forces historians to precede the study and introduction of these documents into scientific circulation by manually deciphering them – a very labor-intensive process. The use of information technology makes it possible to automate this process. The work is aimed at creating a system for automatic recognition of books of decisions of volost courts based on a hybrid approach: neural network methods for recognizing handwritten text and linguistic analysis methods for correcting recognition errors. The paper examines the development of one of the system modules: a letter recognizer. An ensemble approach based on convolutional neural networks is used, which allows achieving an F1-score of 75.1%. The first paragraph – the introduction – sets out the problem, the second provides an analysis of the applicability of existing recognition systems to this problem, and the third describes the development of a specialized system for recognizing books of decisions of volost courts.

KW - оптическое распознавание символов

KW - исторические документы

KW - нейронные сети

KW - волостные суды

KW - извлечение текста

KW - optical character recognition

KW - historical documents

KW - neural networks

KW - volost courts

KW - text extraction

UR - https://www.scopus.com/pages/publications/105031769826

U2 - 10.1109/APEIE66761.2025.11289354

DO - 10.1109/APEIE66761.2025.11289354

M3 - Paper

SP - 1

EP - 4

T2 - 2025 IEEE XVII International Scientific and Technical Conference on Actual Problems of Electronic Instrument Engineering (APEIE)

Y2 - 14 November 2025 through 16 November 2025

ER -