Greek Ancient Majuscule (spaced)

This model is intended for ancient Greek Majuscule manuscripts. The initial data set consisted of three manuscripts of the Old and New Testament. These were the entirety of Codex Bezae Cantabrigiensis (Cambridge University Library Manuscript Nn.2.41) and Bibliotheque nationale de France Manuscript Grec 17 as well as the text of I Samuel - II Kings from Codex Alexandrinus (British Library Royal Manuscript 1.D.V). Though note that, due to access restrictions, images used for the text of Codex Alexandrinus were taken from the 1881 facsimile edition published by the British Library. Since this model was built specifically to handle majuscule manuscripts, Greek capital letters were used to represent the characters. The model was trained only to extract text and no other features. The model was not trained to represent punctuation, accents, breathings or any diacritical marks. Nomina sacra and numerals represented as written (e.g., ΙΗΛ not resolved to ΙΣΡΑΗΛ). However, ligatures and other abbreviations are resolved (e.g., abbreviations of ΚΑΙ, Ν written as a high horizontal line). The model was trained on transcriptions with spaces included between words. Thus, this model will produce spaced transcriptions (or attempt to do so) even for manuscripts written in scripta continua.