A research team from McGill University, Mila – Quebec AI Institute and Facebook AI proposes novel metrics and perturbation functions to detect, quantify and compare trade-offs between robustness and faithfulness in NMT systems, both on the corpus level and with particular examples.

Here is a quick read: Facebook AI, McGill U & Mila Promote ‘Translationese’ to Boost NMT System Faithfulness.

The paper Sometimes We Want Translationese is on arXiv.



Source link