Should we be afraid that neural machine translation (MT) will disrupt — or destroy — our industry? Progress recently seen with neural MT opens new perspectives. Until now, MT has been perceived either as a replacement for human translation, when volumes were too big (such as user generated content or support), or as a cost-killing constraint imposed on translators. Neural technologies bring a set of high-level features — generalization, understanding and global overview — that allow new ways to augment translators and more generally users in need of cross-lingual capabilities. We will share a vision and roadmap for the future for language-aware applications, initial feedback from actual experience and a case study on a qualitative comparison of translation output across different languages and content that will reveal insightful research results, weighing up the relative pros and cons of the different approaches to MT.
Neural MT: Revolution or Evolution