Statistical and Neural Machine Translation Systems of English to Manipuri: A Preliminary Study

The present work reports the findings of the experimental result of English to Manipuri machine translation systems using neural and statistical approaches. The experiment on the resource constrained language pair between the two technological epitome of the machine translation is carried out as a preliminary study using TDIL and Vikaspedia dataset. The finding of the experiment is that the neural based system shows higher BLEU score than the phrase based statistical machine translation system. Human judgement on the translated output for adequacy and fluency reveals interesting facts between the two approaches.