Vladimir Polomac

The paper describes the process of creating a model for the automatic rec- ognition of Serbian Church Slavonic printed books from Venice (from Božidar and Vincenzo Vuković’s printery) by using the Transkribus software platform, based on the principles of artificial intelligence and machine learning. By using the example of Prayer Book (Euchologion) (1538–1540) from Božidar Vuković’s printery, it has been shown that a successful model for the automatic recognition of individual books (with around 5% of unrecognized characters) can also be trained on the material consisting of approximately 4000 words, and that the increased amount of training material (in our case around 38000 words) leads to the improvement of the model and reduced error rate (between 1–2% of unrecognized characters). The most notable result of the paper is manifested through the creation of a generic model for the automatic text recognition of Serbian Church Slavonic books from Božidar and Vincenzo Vuković’s printery. The ini- tial version of the generic model (called Dionisio 1.0. by the Božidar Vuković’s Italian pseudonym – Dionisio della Vecchia) is the first resource for the automatic recognition of the Serbian medieval Cyrillic script, publicly available to all users of the Transkribus software platform (see https://readcoop.eu/model/dionisio-1-0/).

The paper defines the elementary principles for creating an electronic corpus of Serbian medieval charters and letters. The commitment to the principle of maximum representativeness of the corpus of medieval charters and letters, determined entirely by the preserved written legacy (based on manuscripts, microfilms or photographs), excludes the indispensability of applying the principle of balance, while simultaneously satisfying the principle of reliability, since charters and letters known solely by the edition are not included in the corpus. The selection of texts is done according to the diplomatic criterion by excluding the transcripts and copies of documents already available in the original, as well as later transcripts, chronologically and linguistically distant from the assumed original. This approach to the selection of texts is justified by the size of the corpus, as well as by the exceptional cultural and historical significance of medieval charters and letters. The definition of the metadata about corpus texts is determined by their general diplomatic properties, as well as the corpus search needs for diatopic, diachronic and genre variations. Conversion of texts into electronic form strives for fidelity to the original, encompassing the preservation of abbreviations, superscript letters and original punctuation, as well as the absence of accent marks and contemporary rules of capitalization.