Sequence to Sequence Neural Machine translation from Danish to English analysis of ANKI and European Parliament corpus

This paper is an empirical study of the two common corpora for Neural Machine Translation, Anki and European Parliament corpus. Our focus was on Danish to English translation by applying state-of-the-art tools for translation. During this process, we found several challenges, such as AVX and TensorFlow versions; long complex morphology that cannot capture accurate translation. The Danish language itself is a challenge because of its confounded morphology. We then present our modified Attention Alignment Score function, which is showing promising result in Danish to English Neural Machine Translation.