Diacritical Language OCR Based on Neural Network: Case of Amazigh Language

Document paper conversion into electronic format has become indispensable task in many areas, especially for digitizing and translating printed texts. In this context, several approaches have been studied focusing mainly on character recognition for diacritic-free languages. However in this paper, we are interested in the Amazigh language transcribed in Latin, distinguished by its diacritical characters. Thus, we propose to use a system based on neural networks, and to study its behavior against this type of characters.