Alert button

Growing Together: Modeling Human Language Learning With n-Best Multi-Checkpoint Machine Translation

Jun 07, 2020
El Moatez Billah Nagoudi, Muhammad Abdul-Mageed, Hasan Cavusoglu

Figure 1 for Growing Together: Modeling Human Language Learning With n-Best Multi-Checkpoint Machine Translation
Figure 2 for Growing Together: Modeling Human Language Learning With n-Best Multi-Checkpoint Machine Translation
Figure 3 for Growing Together: Modeling Human Language Learning With n-Best Multi-Checkpoint Machine Translation
Figure 4 for Growing Together: Modeling Human Language Learning With n-Best Multi-Checkpoint Machine Translation

Share this with someone who'll enjoy it:

We describe our submission to the 2020 Duolingo Shared Task on Simultaneous Translation And Paraphrase for Language Education (STAPLE) (Mayhew et al., 2020). We view MT models at various training stages (i.e., checkpoints) as human learners at different levels. Hence, we employ an ensemble of multi-checkpoints from the same model to generate translation sequences with various levels of fluency. From each checkpoint, for our best model, we sample n-Best sequences (n=10) with a beam width =100. We achieve 37.57 macro F1 with a 6 checkpoint model ensemble on the official English to Portuguese shared task test data, outperforming a baseline Amazon translation system of 21.30 macro F1 and ultimately demonstrating the utility of our intuitive method.

* Accepted to the 4th Workshop on Neural Generation and Translation (Duolingo Shared Task on Simultaneous Translation And Paraphrase for Language Education Mayhew et al., 2020) collocated with ACL 2020  
View paper onarxiv icon

Share this with someone who'll enjoy it: