Machine Translation is one of the major problems in the field of Natural Language Processing. Over the course, many approaches have been applied in order to solve this problem ranging from traditional rule-based approach, statistical methods to the more recent neural network based methods. Neural network based methods have produced comparable results to that of the existing phrase based model and in some language pairs they have even outperformed the latter. A huge amount of parallel corpus is required for both the SMT and NMT models in order to produce a reasonable result. For some language pairs this data is readily available but for others it may not be the case. This research focuses on the comparative study of how SMT and NMT based machine translation model perform and compare to each other in case where the language pair is under resourced in terms of the availability of parallel corpus.