Statistical Machine Translation

Translation of natural language text using statistical machine translation (SMT) is a supervised machine learning problem. SMT algorithms are trained to learn how to translate by providing many translations produced by human language experts. The field SMT has gained momentum in recent three decades. New techniques are constantly introduced by the researchers. This is survey paper presenting an introduction of the recent developments in the field. The paper also describes the recent research for word alignment and language modelling problems in the translation process. An overview of these two sub problems is enlisted. Along the way, some challenges in machine translation are presented.