Selective high-throughput digitization of large library collections

Large-scale digitization of library collections presents a dilemma: planetary scanning leaves the book undamaged, yet it remains slow and inefficient when millions of items must be processed. The authors examine an alternative approach — selective high-throughput digitization, in which a duplicate copy is selected, its spine is removed, and the resulting sheets are processed on a high-performance sheet-fed scanner. The authors compare this approach with traditional page-by-page scanning and show under what conditions disbinding a copy is justified and when it is inadmissible. Modelling shows that for a collection of one million volumes the imagecapture stage can be reduced from approximately 21 years to between one year and seven months, depending on scanner class — a 21- to 35-fold gain. The study also focuses on item-level accounting and the coordination of digitization. When several libraries independently digitize the same titles, duplicate digital copies are produced. The National Electronic Library (NEL), operating as a federal state information system, already provides for the storage of digital copies and maintains the union electronic catalogue of Russian library collections. However, like other union catalogues, it offers record-level bibliographic control, whereas selective highthroughput digitization requires data on the physical condition of an individual item and its duplicate status. The authors therefore propose not a new national system, but an item-level accounting and digitization coordination subsystem within the existing NEL infrastructure. Its tasks would include identifying duplicate editions, providing end-to-end identification through universal classifiers and identifiers, recording already digitized copies together with their location data, and, on this basis, compiling the array of publications for industrial-scale scanning. The authors also review international experience, from the China Academic Digital Associative Library and Digital Library of India projects to the class action brought by a group of American authors against Anthropic (Bartz et al. v. Anthropic PBC), in which the court distinguished between converting books into digital form by high-throughput digitization and using them to train artificial intelligence language models. In the light of Federal Law No. 243-FZ of 26 July 2026 "On Supporting the Development of Artificial Intelligence Technologies in the Russian Federation", the authors show how large-scale high-throughput digitization may become a source of data for training domestic AI models without prejudice to the interests of rightsholders.