1 # Motivation and Question
3 - Create a stylesheet for NMTs
4 - neural networks getting pretty good in pure translation
5 - often extra information is needed (e.g. domain, tone)
6 - current translators use stylesheets
10 2) multi source translation systems
11 3) complex attention models
16 \includegraphics[scale=0.5]{img/styleguide-example.png}
20 - How can we transform real life/industry data sets to use in deep learning?
21 - How well do current politeness control mechanism perform on real life data sets?
22 - How well do current politeness control mechanism perform on distant language pairs?
24 - How big is the impact of automated tone labeling in comparison to manual labeling?