Abstract:
The ultimate rewarding goal, for me, for an intelligent system is being able to
communicate seamlessly like human. Although there is great progress in the
field of Machine Translation through Statistical Machine Translation (SMT)
over last few years but SMT systems have become increasingly complex due
to its so many independent components and low translation quality that does
not satisfy users, rendering it extremely difficult to make further advancements. Recently, due to emerging of Neural Machine Translation (NMT)
has given a promising solution to machine translation problem. At the core,
NMT model is deep neural network with billions of neurons to learn directly
the map for conversion of sources sentences to target. NMT is a alot powerful due to it being an end-to-end framework. Its performance is significantly
better than SMT in long range dependencies capturing and generalizing well
to unseen texts. This thesis presents how I used NMT in conversion of tenses
of sentences as traditional Classifier based statistical models although translate source language to target language but in doing so accuracy and fluency
was lost
