MS-UEdin Submission to the WMT2018 APE Shared Task: Dual-Source Transformer for Automatic Post-Editing
This work addresses improving translation quality through post-editing for users of machine translation systems, but it is incremental as it builds on prior shared task data and models.
The paper tackled automatic post-editing for machine translation by implementing dual-source transformer models with parameter sharing, achieving state-of-the-art results in the SMT sub-task with a close second in the NMT sub-task, though it questioned the utility of neural-on-neural APE due to weaker NMT performance.
This paper describes the Microsoft and University of Edinburgh submission to the Automatic Post-editing shared task at WMT2018. Based on training data and systems from the WMT2017 shared task, we re-implement our own models from the last shared task and introduce improvements based on extensive parameter sharing. Next we experiment with our implementation of dual-source transformer models and data selection for the IT domain. Our submissions decisively wins the SMT post-editing sub-task establishing the new state-of-the-art and is a very close second (or equal, 16.46 vs 16.50 TER) in the NMT sub-task. Based on the rather weak results in the NMT sub-task, we hypothesize that neural-on-neural APE might not be actually useful.