Simultaneous translation is widely useful but remains one of the most difficult tasks in NLP. Previous work either uses fixed-latency policies, or train a complicated two-staged model using reinforcement learning. We propose a much simpler single model that adds a `delay' token to the target vocabulary, and design a restricted dynamic oracle to greatly simplify training. Experiments on Chinese<->English simultaneous translation show that our work leads to flexible policies that achieve better BLEU scores and lower latencies compared to both fixed and RL-learned policies.
@article{arxiv.1906.01135,
title = {Simultaneous Translation with Flexible Policy via Restricted Imitation Learning},
author = {Baigong Zheng and Renjie Zheng and Mingbo Ma and Liang Huang},
journal= {arXiv preprint arXiv:1906.01135},
year = {2019}
}