Abstract
Machine Translation of text is a fundamental problem in machine learning that resists solutions that do not take into account the dependencies between words and sentences. Recurrent Neural Networks have recently delivered outstanding results in learning about sequential dependencies in many languages. Arabic language as a target language has not received enough attention in the recent language model experiments due to its, structural and semantic difficulties. In this paper, we present a Statistical Machine Translation (SMT) Context Modelling using Recurrent Neural Networks (RNNs) and Latent Dirichlet Allocation (LDA). This research is based on the state-of-the-art RNN language model by Mikolov. Our preliminary contribution is in integrating and presenting a new hybridization to utilize Recurrent Neural Network sequential word learning dependencies as well as Latent Dirichlet Allocation context and topic classification ability to produce the most accurate language scoring.