Optical Character Recognition with a Neural Network Model for Printed Coptic Texts

Furthermore, historical texts are not passed down through the centuries in their entirety but rather contain lacunae and fragmentary words. This makes automatic post-correction more difficult on historical texts than on modern ones. We used two tools to create languageand even documentspecific recognition patterns (or so-called models) to recognize printed Coptic texts. Coptic is the last stage of the pre-Arabic, indigenous Egyptian language. It was used to create a rich and unique body of literature: monastic, “Gnostic,” Manichaean, magical and medical texts, hagiographies, and biblical and patristic translations. We found that Coptic texts have properties which make them excellent candidates for reading by computers. The characters can easily be distinguished due to their limited number and the fact that almost all the hand-written texts exhibit characters with highly consistent forms.