Toward Classification of Arabic Manuscripts Words Based on the Deep Convolutional Neural Networks

Abstract

Deep learning is an area that has seen many developments in recent years. One of these algorithms that have provided good results is Deep Convolutional Neural Networks (DCNN). It is proven to be effective in various fields such as natural language processing, pattern recognition, computer vision, object detection in images, etc. Despite the development of these technologies, Arabic manuscripts in digital libraries still use traditional indexing methods based on metadata, annotation or transcription. In this article, we propose two methods of word classification based on deep learning, the first one uses a simple Neural Network (DNN) and the last one uses a Convolutional Neural Network (DCNN). The idea is to segment words of Arabic manuscripts images and predict the class of each word. The experimental results show the efficient of this classification system based on the DCNN. By comparing the results obtained, we can observe that the DCNN method provides excellent results than those obtained with the DNN method.