Osaka University MT Systems for WAT 2018: Rewarding, Preordering, and Domain Adaptation
Yuki Kawara, Yuto Takebayashi, Chenhui Chu, Yuki Arase · Institutional Repositories DataBase (IRDB) · 2018
In this paper, we present Osaka University MT systems submitted to WAT 2018 shared translation tasks and analysis of their performances.For the ASPEC Japanese-English task, we use our rewarding model on neural machine translation (NMT) and preordering model on phrase-based statistical machine translation (PBSMT).For the Myanmar-English task, we further apply our mixed fine tuning method for domain adaptation on NMT.We report the translation results on these two tasks, where the rewarding model performs the best.