Text Representation for Nonconcatenative Morphology

Abstract

The last six years have seen the immense improvement of the NMT in terms of translation quality. With the help of the neural networks, the NMT has been able to achieve the state-of-the-art results in transla- tion quality. However, the NMT is still not able to achieve translation quality near human levels. In this thesis, we propose new approaches to improve the language representation as input to the NMT. This can be achieved by exploiting language specific knowledge, such as phonetic alterations, the morphology, and the syntax. We p Show more