Simultaneous paraphrasing and translation by fine-tuning Transformer models

This paper describes the third place submission to the shared task on\nsimultaneous translation and paraphrasing for language education at the 4th\nworkshop on Neural Generation and Translation (WNGT) for ACL 2020. The final\nsystem leverages pre-trained translation models and uses a Transformer\narchitecture combined with an oversampling strategy to achieve a competitive\nperformance. This system significantly outperforms the baseline on Hungarian\n(27% absolute improvement in Weighted Macro F1 score) and Portuguese (33%\nabsolute improvement) languages.\n