Design and Implementation of a deep learning model for an automatic translation sequences between English and Arabic

Abstract

Since the turn of the century, machine translation has experienced considerable growth on the Web, with several online systems capable of automatically translating increasingly long Web pages or texts in a matter of seconds. It is a much appreciated help from the general public. Globalization has had considerable effects on the development of the language industry, and more particularly in machine translation where demand continues to grow. Thus, the need for more reliable machine translation systems is increasing . That is why we are interested in designing a system for machine translation (translating text from one language to another, in our case from English to Arabic), and in order to implement this process, we used a modern adaptation of machine translation: neural machine translation. This approach depends largely extent on the neurons in our brain and how we learn as humans. We proposed an approach to create a translation system between English and Arabic , and then test it on a set of sentences, paragraphs and short texts. In order to achieve this, we built a model that processes and translates some English sequence as a sequence of words , producing sentences that are also made up of a sequence of Arabic words . More advanced MT models usually process sentences word for by word . This method used is based on the encoding / decoding model in which both the encoder and decoder units are Long Short Term Memory models, which in turn are considered as type of Recurrent Neural Networks and the latter is also classified as a type of Neural Networks Neural Machine Translation relies on Neural Networks that are part of Natural Language Processing and implicit in Deep Learning .According to the results we evaluated by Bleu score, the translation was close to the average.