Measuring Performance of NMT for Attention Model and Sequence to Sequence Model without Attention using BLEU

Abstract

An abundant amount of information is available online in English languages. This information is transformed into native languages so that local people can understand it. These local individuals are not much aware of the English language. To convert any text to different languages without the involvement of human-machine translation is a very efficient technique. Anyone who thinks to translate any language to another language manually is a very tedious and time-consuming task. Among all the machine translation technique which are available in recent time, NMT is the most efficient translation technique. In this paper, the translation of English-French is done with the help of the attention model and without the attention model. I have used the BLEU score to estimate the performance of the model. BLEU is a parameterized metric, in which when we change the parameter then their value is changed largely.