Možnosti aplikácie metódy digitálnej transkripcie historických rukopisných textov pri sprístupňovaní archívnych fondov

Nagy, Imrich, 1972- Možnosti aplikácie metódy digitálnej transkripcie historických rukopisných textov pri sprístupňovaní archívnych fondov = The Possibilities of application the method of digital transcription of historical manuscript texts in the process of accessing the archival fonds / Imrich Nagy. -- Modern technologies that use elements of the artificial intelligence based on neural engine, offer new possibilities in accessing the historical manuscript texts. So it is the Transkribus platform, developed within the European project Horizon 2020 READ. To verify its functionality, the authors chose a contemporary archival aid – a catalog of correspondence of Koháry family, processed by J. Csákós from the fonds of the State Archives in Banská Bystrica. The numerical catalog offers abstracts and partly transcripts to 6632 letters on 4140 sheets of A3 paper format. The catalog was digitized by scanning with a camera of an iPhone 11 Pro and ScanTent with a resolution of 192 DPI, sufficient for the HTR+ method using Transkribus platform. On a sample of 29 images containing the first 53 pages of the Csákós’catalog, the basic model for the automatic trascription was practised. The success of the model is determined statistically by the CER indicator which marks the ratio of erroneous characters in the automatically generated transcript. The model achieved CER 4,11 % on the verification file pages. In general, a model with CER ≤ 10 % is rated as functional and with CER ≤ 5 % as succesful. Based on the authors’model, they performed an automatic transcription of another 28 images containing the next 50 pages of Csákós’catalog. The average CER value for automatically transcribed pages was 5,26 %. As a part of the experiment, they expanded the sample file for training the model with corrected pages from the automatic transcription. The corrected model achieved CER 3,08 %. Such a model is fully functional for the automatic transcription. On its base it is possible to transcribe the complete catalog of Koháry’s correspondence. Transcription out puts will allow further research of correspondence: identification of persons, locations, events, dating and other data.

In Slovenská archivistika. -- Bratislava : Ministerstvo vnútra Slovenskej republiky, 2021. -- ISSN 0231-6722. -- Roč. 51, č. 2 (2021), s. 53-67

1. digitalizácia 2. transkripcia 3. archívne fondy 4. články

I. Slovenská archivistika. -- Roč. 51, č. 2 (2021), s. 53-67
BB301