Complementary Approaches Built as Web Service for Arabic Handwriting OCR Systems via Amazon Elastic MapReduce (EMR) Model

Arabic Optical Character Recognition (OCR) as Web Services represents a major challenge for handwritten document recognition. A variety of approaches, methods, algorithms and techniques have been proposed in order to build powerful Arabic OCR web services. Unfortunately, these methods could not succeed in achieving this mission in case of large quantity Arabic handwritten documents. Intensive experiments and observations revealed that some of the existing approaches and techniques are complementary and can be combined to improve the recognition rate. Designing and implementing these recent sophisticated complementary approaches and techniques as web services are commonly complex; they require strong computing power to reach an acceptable recognition speed especially in case of large quantity documents. One of the possible solutions to overcome this problem is to benefit from distributed computing architectures such as cloud computing. This paper describes the design and implementation of Arabic Handwriting Recognition as a web service (AHRweb service) based on the complementary approach K-Nearest Neighbor (KNN) /Support Vector Machine (SVM) (K-NN/SVM) via Amazon Elastic Map Reduce (EMR) model. The experiments were conducted on a cloud computing environment with a real large scale handwriting dataset from the Institut Für Nachrichtentechnik (IFN)/ Ecole Nationale d'Ingénieur de Tunis (ENIT) IFN/ENIT database. The J-Sim (Java Simulator) was used as a tool to generate and analyze statistical results. Experimental results show that Amazon Elastic Map Reduce (EMR) model constitutes a very promising framework for enhancing large Arabic Handwriting Recognition (AHR) web service performances