ABSTRACT
Every change in the paradigm of machine translation (MT) architectures has lifted the potential of MT technologies, evoking expectations, or dreams, of ‘human parity’ in general-purpose MT. Recent years have witnessed a significant improvement in MT performance due to the advent of neural MT, which makes use of a huge volume of text data to train a deep neural network model to generate target text from source text in a single, unified process. An increasing number of companies, governments and individuals have started to use MT tools to meet their various needs, many of which might not have emerged without the development of MT. Technologies have been expanding the sphere of translation market more than ever before.
