ABSTRACT
Machine Translation is the automated method of conversion text from one language to other foreign languages by using intelligence approach of automated algorithms, when machine learning concepts applied to it. Over the conversion process, the mechanism utilized with intermediate process to translate from one source language to a foreign language. It is the part of artificial intelligence (AI), when intelligence has the prediction capability makes as machine learning and much accuracy, known as deep learning. Conversion of one form to another form is called translation that makes trending as a world as a single language, that form of automated conversions called machine translation. This kind of translation requires 54the background knowledge about the both languages or for multiple languages, in this way a classical machine translation needs to satisfies a set of constraints based as guidelines. Because of some languages produces same text over different meanings over part of utilization in the context and, sometimes provides same meaning over different words. So, the basic two methods of classification for machine translations were statistical and neural machine translations (NMTs). In the traditional method, usage of statistical machine translation (SMT) approach to perform translation was the way to predict possible best outcome with definite algorithms. But in NMT, approach applies the dynamic algorithms for best predictability of word on translation according to the context appropriately. In the chapter, the proposed assertion explains about all form of NMT with complete translation of process involves neural network, that produces more number of accessible input to find the best possible output according to utilization. This kind of translation applies multiple strategies on different stages for translation, Stage one implements the translated based on word with complete sentence and, Stage 2 implements the translated on model over the word within the sentence context. The solidity of NMT allows the learning ability over point-to-point bases on the background knowledge input to the predicable target output. And also brief discussion on various kinds of NMT conversion principles such as like Google neural machine translation (GNMT), open source neural machine translation (OpenNMT), deep neural machine translation (DeepNMT), and so on.
